Transcript Geant4 – Updates Event biasing Cuts per region

June 2005, Geant4 v7.0p01
More on kernel
Makoto Asai (SLAC)
Geant4 Tutorial Course
the 2nd Finnish Geant4 Workshop
June 6-7 2005, Helsinki Institute of Physics
Contents







Detector sensitivity
Track and step statuses
Attaching user information to G4 classes
Stacking mechanism
Cuts per region
Event biasing
Fast simulation (Shower parameterization)
More on Kernel - M.Asai (SLAC)
2
Detector Sensitivity
Sensitive detector and Hit

Each Logical Volume can have a pointer to a sensitive detector.


Then this volume becomes sensitive.
Hit is a snapshot of the physical interaction of a track or an accumulation of
interactions of tracks in the sensitive region of your detector.

A sensitive detector creates hit(s) using the information given in G4Step
object. The user has to provide his/her own implementation of the detector
response.


UserSteppingAction class should NOT do this.
Hit objects, which are still the user’s class objects, are collected in a G4Event
object at the end of an event.
More on Kernel - M.Asai (SLAC)
4
Detector sensitivity

A sensitive detector either

constructs one or more hit objects or

accumulates values to existing hits
using information given in a G4Step object.

Note that you must get the volume information from the “PreStepPoint”.
Boundary
Step
End of step point
Begin of step point
More on Kernel - M.Asai (SLAC)
5
Digitizer module and digit

Digit represents a detector output (e.g. ADC/TDC count, trigger signal, etc.).

Digit is created with one or more hits and/or other digits by a user's concrete
implementation derived from G4VDigitizerModule.

In contradiction to the sensitive detector which is accessed at tracking time
automatically, the digitize() method of each G4VDigitizerModule must be
explicitly invoked by the user’s code (e.g. at EventAction).
More on Kernel - M.Asai (SLAC)
6
Hit class



Hit is a user-defined class derived from G4VHit.
You can store various types information by implementing your own concrete Hit class.
For example:

Position and time of the step

Momentum and energy of the track

Energy deposition of the step

Geometrical information

or any combination of above
Hit objects of a concrete hit class must be stored in a dedicated collection which is
instantiated from G4THitsCollection template class.

The collection will be associated to a G4Event object via G4HCofThisEvent.

Hits collections are accessible

through G4Event at the end of event.


to be used for analyzing an event
through G4SDManager during processing an event.

to be used for event filtering.
More on Kernel - M.Asai (SLAC)
7
Implementation of Hit class
#include "G4VHit.hh"
class MyDriftChamberHit : public G4VHit
{
public:
MyDriftChamberHit(some_arguments);
virtual ~MyDriftChamberHit();
virtual void Draw();
virtual void Print();
private:
// some data members
public:
// some set/get methods
};
#include “G4THitsCollection.hh”
#typedef G4THitsCollection<MyDriftChamberHit>
MyDriftChamberHitsCollection;
More on Kernel - M.Asai (SLAC)
8
Sensitive Detector class

Sensitive detector is a user-defined class derived from G4VSensitiveDetector.
#include "G4VSensitiveDetector.hh"
#include "MyDriftChamberHit.hh"
class G4Step;
class G4HCofThisEvent;
class MyDriftChamber : public G4VSensitiveDetector
{
public:
MyDriftChamber(G4String name);
virtual ~MyDriftChamber();
virtual void Initialize(G4HCofThisEvent*HCE);
virtual G4bool ProcessHits(G4Step*aStep,
G4TouchableHistory*ROhist);
virtual void EndOfEvent(G4HCofThisEvent*HCE);
private:
MyDriftChamberHitsCollection * hitsCollection;
G4int collectionID;
};
More on Kernel - M.Asai (SLAC)
9
Sensitive detector

A tracker detector typically generates a hit for every single step of every single
(charged) track.



Position and time

Energy deposition of the step

Track ID
A calorimeter detector typically generates a hit for every cell, and accumulates
energy deposition in that cell for all steps of all tracks.


A tracker hit typically contains
A calorimeter hit typically contains

Sum of deposited energy

Cell ID
You can instantiate more than one objects for one sensitive detector class. Each
object should have its unique detector name.

For example, each of two sets of drift chambers can have their dedicated
sensitive detector objects. But, the functionalities of them are exactly the
same to each other and thus they can share the same class. See
examples/extended/analysis/A01 as an example.
More on Kernel - M.Asai (SLAC)
10
Implementation of Sensitive Detector - 1
MyDriftChamber::MyDriftChamber(G4String detector_name)
:G4VSensitiveDetector(detector_name),
collectionID(-1)
{
collectionName.insert(“collection_name");
}

In the constructor, define the name of the hits collection which is handled by
this sensitive detector

In case your sensitive detector generates more than one kinds of hits (e.g.
anode and cathode hits separately), define all collection names.
More on Kernel - M.Asai (SLAC)
11
Implementation of Sensitive Detector - 2
void MyDriftChamber::Initialize(G4HCofThisEvent*HCE)
{
if(collectionID<0) collectionID = GetCollectionID(0);
hitsCollection = new MyDriftChamberHitsCollection
(SensitiveDetectorName,collectionName[0]);
HCE->AddHitsCollection(collectionID,hitsCollection);
}




Initialize() method is invoked at the beginning of each event.
Get the unique ID number for this collection.
 GetCollectionID() is a heavy operation. It should not be used for every events.
 GetCollectionID() is available after this sensitive detector object is registered
to G4SDManager. Thus, this method cannot be used in the constructor of this
detector class.
Instantiate hits collection(s) and attach it/them to G4HCofThisEvent object given
in the argument.
In case of calorimeter-type detector, you may also want to instantiate hits for all
calorimeter cells with zero energy depositions, and insert them to the collection.
More on Kernel - M.Asai (SLAC)
12
Implementation of Sensitive Detector - 3
G4bool MyDriftChamber::ProcessHits
(G4Step*aStep,G4TouchableHistory*ROhist)
{
MyDriftChamberHit* aHit = new MyDriftChamberHit();
...
// some set methods
...
hitsCollection->insert(aHit);
return true;
}




This ProcessHits() method is invoked for every steps in the volume(s) where this
sensitive detector is assigned.
In this method, generate a hit corresponding to the current step (for tracking
detector), or accumulate the energy deposition of the current step to the existing
hit object where the current step belongs to (for calorimeter detector).
Don’t forget to collect geometry information (e.g. copy number) from
“PreStepPoint”.
Currently, returning boolean value is not used.
More on Kernel - M.Asai (SLAC)
13
Implementation of Sensitive Detector - 4
void MyDriftChamber::EndOfEvent(G4HCofThisEvent*HCE)
{;}

This method is invoked at the end of processing an event.

It is invoked even if the event is aborted.

It is invoked before UserEndOfEventAction.
More on Kernel - M.Asai (SLAC)
14
Touchable




As mentioned already, G4Step has two G4StepPoint objects as its
starting and ending points. All the geometrical information of the
particular step should be taken from “PreStepPoint”.
 Geometrical information associated with G4Track is basically same as
“PostStepPoint”.
Each G4StepPoint object has
 Position in world coordinate system
 Global and local time
 Material
 G4TouchableHistory for geometrical information
G4TouchableHistory object is a vector of information for each
geometrical hierarchy.
 copy number
 transformation / rotation to its mother
Since release 4.0, handles (or smart-pointers) to touchables are
intrinsically used. Touchables are reference counted.
More on Kernel - M.Asai (SLAC)
15
Copy number

Suppose a calorimeter is made of
4x5 cells.



In reality, there is only one physical
volume object for each level. Its
position is parameterized by its
copy number.
To get the copy number of each
level, suppose what happens if a
step belongs to two cells.



and it is implemented by two
levels of replica.
0
1 CopyNo
2 =0 3
4
0
1 CopyNo
2 =1 3
4
0
1 CopyNo
2 =2 3
4
0
1 CopyNo
2 =3 3
4
Remember geometrical information in G4Track is identical to
"PostStepPoint".
You cannot get the collect copy number for "PreStepPoint" if you directly
access to the physical volume.
Use touchable to get the proper copy number, transform matrix, etc.
More on Kernel - M.Asai (SLAC)
16
Touchable

G4TouchableHistory has information of geometrical hierarchy of the point.
G4Step* aStep;
G4StepPoint* preStepPoint = aStep->GetPreStepPoint();
G4TouchableHistory* theTouchable =
(G4TouchableHistory*)(preStepPoint->GetTouchable());
G4int copyNo = theTouchable->GetVolume()->GetCopyNo();
G4int motherCopyNo
= theTouchable->GetVolume(1)->GetCopyNo();
G4int grandMotherCopyNo
= theTouchable->GetVolume(2)->GetCopyNo();
G4ThreeVector worldPos = preStepPoint->GetPosition();
G4ThreeVector localPos = theTouchable->GetHistory()
->GetTopTransform().TransformPoint(worldPos);
More on Kernel - M.Asai (SLAC)
17
Readout geometry



In some cases of most complicated geometries, it is not easy to define volume
boundaries corresponding to the readout segmentation.
Readout geometry is a virtual and artificial geometry which can be defined in
parallel to the real detector geometry.
Readout geometry is optional. May have more than one.


Each one should be associated to a sensitive detector.
Note that a step is not limited by the boundary of readout geometry.
More on Kernel - M.Asai (SLAC)
18
Defining a sensitive detector

Basic strategy
G4LogicalVolume* myLogCalor = ……;
G4VSensetiveDetector* pSensetivePart =
new MyCalorimeter(“/mydet/calorimeter1”);
G4SDManager* SDMan = G4SDManager::GetSDMpointer();
SDMan->AddNewDetector(pSensitivePart);
myLogCalor->SetSensitiveDetector(pSensetivePart);

Each detector object must have a unique name.



Some logical volumes can share one detector object.
More than one detector objects can be instantiated from one detector class
with different detector name.
One logical volume cannot have more than one detector objects. But, one
detector object can generate more than one kinds of hits.

e.g. a drift chamber class may generate anode and cathode hits
separately.
More on Kernel - M.Asai (SLAC)
19
G4HCofThisEvent

A G4Event object has a G4HCofThisEvent object at the end of (successful) event
processing. G4HCofThisEvent object stores all hits collections made within the
event.

Pointer(s) may be NULL if collection(s) are not created in the particular
event.

Hits collections are stored by pointers of G4VHitsCollection base class. Thus,
you have to cast them to types of individual concrete classes.

The index number of a Hits collection is unique and stable for a run and can
be obtained by
G4SDManager::GetCollectionID(“detName/colName”);

The index table is also stored in G4Run.
More on Kernel - M.Asai (SLAC)
20
Usage of G4HCofThisEvent
static int CHCID = -1;
if(CHCID<0) G4SDManager::GetSDMpointer()
->GetCollectionID("myDet/calorimeter1/collection1");
G4HCofThisEvent* HCE = evt->GetHCofThisEvent();
MyCalorimeterHitsCollection* CHC = 0;
cast
if(HCE)
{CHC = (MyCalorimeterHitsCollection*)(HCE->GetHC(CHCID));}
if(CHC)
{ int n_hit = CHC->entries();
G4cout<<"Calorimeter has ”<<n_hit<<" hits."<<G4endl;
for(int i1=0;i1<n_hit;i1++)
{ MyCalorimeterHit* aHit = (*CHC)[i1];
aHit->Print(); }
}

This scheme can be adapted also for Digitization.
More on Kernel - M.Asai (SLAC)
21
Track and Step Statuses







Track
status
At the end of each step, according to the processes involved, the state of a track
may be changed.
 The user can also change the status in UserSteppingAction.
 Statuses shown in yellow are artificial, i.e. Geant4 kernel won’t set them,
but the user can set.
fAlive
 Continue the tracking.
fStopButAlive
 The track has come to zero kinetic energy, but still AtRest process to occur.
fStopAndKill
 The track has lost its identity because it has decayed, interacted or gone
beyond the world boundary.
 Secondaries will be pushed to the stack.
fKillTrackAndSecondaries
 Kill the current track and also associated secondaries.
fSuspend
 Suspend processing of the current track and push it and its secondaries to
the stack.
fPostponeToNextEvent
 Postpone processing of the current track to the next event.
 Secondaries are still being processed within the current event.
More on Kernel - M.Asai (SLAC)
24
Set the track status

In UserSteppingAction, user can change the status of a track.
void MySteppingAction::UserSteppingAction
(const G4Step * theStep)
{
G4Track* theTrack = theStep->GetTrack();
if(…) theTrack->SetTrackStatus(fSuspend);
}

If a track is killed, physics quantities of the track (energy, charge, etc.)
are not conserved but completely lost.
More on Kernel - M.Asai (SLAC)
25
Step status









Step status is attached to G4StepPoint to indicate why that particular step was
determined.
 Use “PostStepPoint” to get the status of this step.
 “PreStepPoint” has the status of the previous step.
fWorldBoundary
 Step reached the world boundary
fGeomBoundary
 Step is limited by a geometry boundary
fAtRestDoItProc, fAlongStepDoItProc, fPostStepDoItProc
 Step is limited by a AtRest, AlongStep or PostStep process
fUserDefinedLimit
 Step is limited by the user Step limit in the logical volume
fExclusivelyForcedProc
 Step is limited by an exclusively forced PostStep process
fUndefined
 Step not defined yet
If you want to identify the first step in a volume, pick fGeomBoudary status in
PreStepPoint.
If you want to identify a step getting out of a volume, pick fGeomBoundary
status in PostStepPoint
More on Kernel - M.Asai (SLAC)
26
Attaching user information to
some kernel classes
Attaching user information


Abstract classes

User can use his/her own class derived from the provided base class

G4Run, G4VHit, G4VDigit, G4VTrajectory, G4VTrajectoryPoint
Concrete classes

User can attach a user information class object

G4Event - G4VUserEventInformation

G4Track - G4VUserTrackInformation

G4PrimaryVertex - G4VUserPrimaryVertexInformation

G4PrimaryParticle - G4VUserPrimaryParticleInformation

G4Region - G4VUserRegionInformation

User information class object is deleted when associated Geant4 class
object is deleted.
More on Kernel - M.Asai (SLAC)
28
Trajectory and trajectory point

Trajectory and trajectory point class objects persist until the end of an
event.


G4VTrajectory is the abstract base class to represent a trajectory, and
G4VTrajectoryPoint is the abstract base class to represent a point which
makes up the trajectory.


And most likely stored to disk as "simulation truth"
In general, trajectory class is expected to have a vector of trajectory
points.
Geant4 provides G4Trajectoy and G4TrajectoryPoint concrete classes as
defaults. These classes keep only the most common quantities.

If the user wants to keep some additional information and/or wants
to change the drawing style of a trajectory, he/she is encouraged to
implement his/her own concrete classes.
More on Kernel - M.Asai (SLAC)
29
Creation of trajectories


Naïve creation of trajectories occasionally causes a memory
consumption concern, especially for high energy EM showers.
In UserTrackingAction, you can switch on/off the creation of a trajectory
for the particular track.
void MyTrackingAction
::PreUserTrackingAction(const G4Track* aTrack)
{
if(...)
{ fpTrackingManager->SetStoreTrajectory(true); }
else
{ fpTrackingManager->SetStoreTrajectory(false); }
}

If you want to use user-defined trajectory, object should be instantiated
in this method and set to G4TrackingManager by SetTrajectory() method.
More on Kernel - M.Asai (SLAC)
30
Bookkeeping issues

Connection from G4PrimaryParticle to G4Track
G4int G4PrimaryParticle::GetTrackID()

Returns the track ID if this primary particle had been converted into G4Track,
otherwise -1.


Both for primaries and pre-assigned decay products
Connection from G4Track to G4PrimaryParticle
G4PrimaryParticle* G4DynamicParticle::GetPrimaryParticle()

Returns the pointer of G4PrimaryParticle object if this track was defined as a
primary or a pre-assigned decay product, otherwise null.

G4VUserPrimaryVertexInformation, G4VUserPrimaryParticleInformation and
G4VUserTrackInformation can be utilized for storing additional information.

Information in UserTrackInformation should be then copied to user-defined
trajectory class, so that such information is kept until the end of the event.
More on Kernel - M.Asai (SLAC)
31
Examples/extended/
runAndEvent/RE01

An example for connecting
G4PrimaryParticle, G4Track, hits
and trajectories, by utilizing
PrimaryTrackID = 1
PrimaryTrackID = 1
SourceTrackID = 1
SourceTrackID = 1
RE01TrackInformation
G4VUserTrackInformation and
PrimaryTrackID = 1
SourceTrackID = 1
G4VUserRegionInformation.

PrimaryTrackID = 1
SourceTrackID = 1
SourceTrackID means the
ID of a track which gets
PrimaryTrackID = 1
SourceTrackID = 4
into calorimeter.

PrimaryTrackID is copied
to UserTrackInformation
PrimaryTrackID = 1
of
SourceTrackID
=3
daughter tracks.

PrimaryTrackID = 1
PrimaryTrackID = 1
SourceTrackID
= 4= 1
PrimaryTrackID
SourceTrackID = 4
SourceTrackID = 4
SourceTrackID is updated for
secondaries born in tracker,
while just copied in calorimeter.
PrimaryTrackID = 2
SourceTrackID = 2
More on Kernel - M.Asai (SLAC)
32
Examples/extended/runAndEvent/RE01
Trajectory of track6782
Tracker hits of track6782
Calorimeter hits of track6782
Energy deposition includes not only
muon itself but also all secondary
EM showers started inside the
calorimeter.
More on Kernel - M.Asai (SLAC)
33
RE01RegionInformation

This example RE01 has three regions, i.e. default world region, tracker region
and calorimeter region.
 Each region has its unique object of RE01RegionInformation class.
class RE01RegionInformation : public G4VUserRegionInformation
{
…
public:
G4bool IsWorld() const;
G4bool IsTracker() const;
G4bool IsCalorimeter() const;
…
};

Through step->preStepPoint->physicalVolume->logicalVolume->region->
regionInformation, you can easily identify in which region the current step
belongs.
 Don’t use volume name to identify.
More on Kernel - M.Asai (SLAC)
34
Cuts per region
Cuts per Region





Geant4 has had a unique production threshold (‘cut’) expressed in
length (i.e. minimum range of secondary).
 For all volumes
 Possibly different for each particle.
Yet appropriate length scales can vary greatly between different areas
of a large detector
 E.g. a vertex detector (5 mm) and a muon detector (2.5 cm).
 Having a unique (low) cut can create a performance penalty.
Requests from ATLAS, BABAR, CMS, LHCb, …, to allow several cuts
 Globally or per particle
New functionality,
 enabling the tuning of production thresholds at the level of a subdetector, i.e. region.
 Cuts are applied only for gamma, electron and positron and only
for processes which have infrared divergence.
‘Full release’ in Geant4 5.1 (end April, 2003)
 Comparable run-time performance
More on Kernel - M.Asai (SLAC)
36
Region

Introducing the concept of region.

Set of geometry volumes, typically
of a sub-system;

barrel + end-caps of the
Default
Region
Region B
calorimeter;

“Deep” areas of support
structures can be a region.


Or any group of volumes;
A set of cuts in range is associated to a
region;

a different range cut for each
particle among gamma, e-, e+ is
Region
B
Region A
Region
B
C
C
Region B
allowed in a region.
More on Kernel - M.Asai (SLAC)
37
Region and cut


Each region has its unique set of cuts.
World volume is recognized as the default
region and the default cuts defined in
Physics list are used for it.


A logical volume becomes a root logical
volume once it is assigned to a region.


User is not allowed to define a region
to the world volume or a cut to the
default region.
World Volume - Default Region
Root logical - Region A
Root logical
- Region B
All daughter volumes belonging to the
root logical volume share the same
region (and cut), unless a daughter
volume itself becomes to another root.
Important restriction :

No logical volume can be shared by
more than one regions, regardless of
root volume or not.
More on Kernel - M.Asai (SLAC)
38
Stack management
Track stacks in Geant4


By default, Geant4 has three track stacks.

"Urgent", "Waiting" and "PostponeToNextEvent"

Each stack is a simple "last-in-first-out" stack.

User can arbitrary increase the number of stacks.
ClassifyNewTrack() method of UserStackingAction decides which stack each
newly storing track to be stacked (or to be killed).



A Track is popped up only from Urgent stack.
Once Urgent stack becomes empty, all tracks in Waiting stack are transferred to
Urgent stack.


By default, all tracks go to Urgent stack.
And NewStage() method of UsetStackingAction is invoked.
Utilizing more than one stacks, user can control the priorities of processing
tracks without paying the overhead of "scanning the highest priority track"
which was the only available way in Geant3.

Proper selection/abortion of tracks/events with well designed stack
management provides significant efficiency increase of the entire simulation.
More on Kernel - M.Asai (SLAC)
40
Stacking mechanism
Temporary
Stack
Urgent
Stack
Transfer
Waiting
Stack
Transfer
Postpone To
Next Event
Stack
Pop
primary
tracks
User Stacking
Action
Classify
NewStage
Prepare Reclassify
Push
PushNew Event
Pop
Push
Push
Stacking
Manager
End Of
Event
Event Manager
Pop
Process
One
Track
secondary
and suspended
tracks
Deleted
Tracking
Manager
RIP
More on Kernel - M.Asai (SLAC)
41
G4UserStackingAction

User has to implement three methods.

G4ClassificationOfNewTrack ClassifyNewTrack(const G4Track*)


Invoked every time a new track is pushed to G4StackManager.

Classification
fUrgent - pushed into Urgent stack

fWaiting - pushed into Waiting stack

fPostpone - pushed into PostponeToNextEvent stack

fKill - killed
void NewStage()




Invoked when Urgent stack becomes empty and all tracks in Waiting stack
are transferred to Urgent stack.
All tracks which have been transferred from Waiting stack to Urgent stack
can be reclassified by invoking stackManager->ReClassify()
void PrepareNewEvent()

Invoked at the beginning of each event for resetting the classification
scheme.
More on Kernel - M.Asai (SLAC)
42
ExN04StackingAction


ExampleN04 has simplified collider
detector geometry and event samples of
Higgs decays into four muons.
Stage 0




Only primary muons are pushed into
Urgent stack and all other primaries
and secondaries are pushed into
Waiting stack.
All of four muons are tracked without
being bothered by EM showers caused
by delta-rays.
Once Urgent stack becomes empty (i.e.
end of stage 0), number of hits in
muon counters are examined.
Proceed to next stage only if sufficient
number of muons passed through
muon counters. Otherwise the event is
aborted.
More on Kernel - M.Asai (SLAC)
43
ExN04StackingAction

Stage 1




Only primary charged particles are
pushed into Urgent stack and all other
primaries and secondaries are pushed
into Waiting stack.
All of primary charged particles are
tracked until they reach to the surface
of calorimeter. Tracks reached to the
calorimeter surface are suspended and
pushed back to Waiting stack.
All charged primaries are tracked in
the tracking region without being
bothered by the showers in calorimeter.
At the end of stage 1, isolation of
muon tracks is examined.
More on Kernel - M.Asai (SLAC)
44
ExN04StackingAction

Stage 2

Only tracks in "region of interest" are
pushed into Urgent stack and all other
tracks are killed.

Showers are calculated only inside of
"region of interest".
More on Kernel - M.Asai (SLAC)
45
Event biasing
Event biasing in Geant4



Event biasing (variance reduction) techniques are a vital requirement for
many applications
These feature could be utilized by many application fields such as

Shielding

Radiation environment assessment

Dosimetry
Since Geant4 is a toolkit and also all source code is open, the user can
do whatever he/she wants.


Capable users in experiments/institutions created their own
implementations of event biasing.
Yet it is more convenient for user if Geant4 itself provides most
commonly used event biasing techniques.
More on Kernel - M.Asai (SLAC)
47
Event biasing techniques







Production cuts / threshold
 This is a biasing technique – most popular for many applications
Geometry based biasing
 Importance weighting for volume/region
 Duplication or sudden death of tracks
Leading particle biasing
 Taking only the most energetic (or most important) secondary
Primary event biasing
 Biasing primary events and/or primary particles in terms of type of
event, momentum distribution, etc.
Forced interaction
 Force a particular interaction, e.g. within a volume
Enhanced process or channel
 Increasing cross section for a process
Physics based biasing
 Biasing secondary production in terms of particle type, momentum
distribution, cross-section, etc.
 Weight on Track / Event
More on Kernel - M.Asai (SLAC)
48
Current features in Geant4



Partial MARS migration
 n, p, pi, K (< 5 GeV)

Since Geant4 0.0

Since Geant4 3.0
General particle source module
 Primary particle biasing
Radioactive decay module
 Physics process biasing in terms of decay products and
momentum distribution




Since Geant4 3.0
Geometry based biasing
 Weight associating with real volume or artificial volume

Since Geant4 4.2

Since Geant4 5.2
Weight cutoff and weight window
Cross-section biasing and leading particle biasing for hadronic
processes

Since Geant4 7.0
More on Kernel - M.Asai (SLAC)
49
Geometrical importance biasing
I = 1.0
W=1.0
I = 2.0
W=0.5
W=0.5
P = 0.5




Define importance for each
geometrical region
Splitting a track,
 Eg creating two particles with
half the ‘weight’ if it moves
into volume with double
importance value.
Russian-roulette in opposite
direction.
Scoring particle flux with weights
 At the surface of volumes
More on Kernel - M.Asai (SLAC)
50
Using importance biasing





Decide whether to bias in “mass” geometry, or in a dedicated ‘parallel’
geometry (“importance geometry”).
Assign an importance value to all volumes in this geometry
 The importance is a number (double)
Register the processes for importance biasing (and scoring, optionally)
to each particle type (eg neutron, gamma, proton, …)
 Examples show easy ways to do this.
Caveats
 World of importance geometry must ‘overlap’ exactly with mass
world
 Biasing and scoring of charged particles in a field is not yet
supported
More details
 Users can choose their importance sampling algorithm,
 Or accept the default one (‘equal weight’).
 For customization & further information see “Geant4 User's
Guide for Application Developers”, Section 3.7
More on Kernel - M.Asai (SLAC)
51
Biasing example B01


Examples demonstrating the use of importance biasing and
scoring can be found in examples/extended/biasing
B01
 Shows the importance sampling and scoring in the mass
(tracking) geometry
 Option to show weight window
 Geometry is 80 cm high concrete cylinder divided into 18
slabs
 Importance values assigned to 18 concrete slabs in
the detector construction - for simplicity.
 The G4Scorer is used for the scoring
 Top level class uses the frame work provided for
scoring.
More on Kernel - M.Asai (SLAC)
52
Example biasing B02




Show how to use
 importance sampling in a parallel geometry and
 a customized scoring making use of the scoring framework.
A simple mass geometry consists of a 180 cm high concrete cylinder
A parallel geometry is created to hold importance values for slabs of
width 10cm and for scoring.
 Note: The parallel world volume must overlap the mass world
volume
 The radii of the slabs is larger than the radius of the concrete
cylinder in the mass geometry.
 The importance value is assigned to each ‘G4GeometryCell’
 To do this, pairs of G4GeometryCell and importance values are
stored in the importance store.
The scoring uses the G4CellSCorer and one customized scorer for the
last slab.
More on Kernel - M.Asai (SLAC)
53
Example B03






Uses Geant4 importance sampling and scoring through python.
It creates a simple histogram.
It demonstrates how to use a customized scorer and importance
sampling in combination with a scripting language, python.
Geant4 code is executed from a python session.
 Note: the swig package is used to create python shadow classes and
to generate the code necessary to use the Geant4 libraries from a
python session.
It can be built and run using the PI implementation of AIDA
 For this see http://cern.ch/PI.
At the end a histogram called "trackentering.hbook" is create.
More on Kernel - M.Asai (SLAC)
54
The Weight Window Technique




The weight window technique is a weight-based algorithm – generally used
together with other techniques as an alternative to importance sampling:
 It applies splitting and Russian roulette depending on space (cells) and
energy
 user defines weight windows in contrast to defining importance values as
in importance sampling
It checks the value of the particle weight
 Comparing it to a ‘window’ of weights defined for the current energyspace cell.
 Doing splitting or roulette in case if it is outside, resulting in 0 or more
particles ‘inside’ the window.
apply in combination with other techniques such as cross-section biasing,
leading particle and implicit capture, or combinations of these.
A weight window may
be specified for every
cell and for several
energy regions: spaceenergy cell .
More on Kernel - M.Asai (SLAC)
55
Leading particle biasing


Simulating a full shower is an
expensive calculation.
Instead of generating a full shower,
trace only the most energetic
secondary.
 Other secondary particles are
immediately killed before being
stacked.
 Convenient way to roughly
estimate, e.g. the thickness of
a shield.
 Of course, physical quantities
such as energy are not
conserved for each event.
More on Kernel - M.Asai (SLAC)
56
Scoring



Scoring is provided by a framework and is done according to particle
type.
 It is possible to score particles of different types into the same scorer.
The framework may also be easily used for customized scoring.
Scoring may be applied to a mass or a parallel geometry. It is done with
an object generically called a scorer using a sampler.
 The scorer receives the information about every step taken by a
particle of chosen type. This information consists a G4Step and a
G4GeometryCellStep (created for scoring and importance sampling).
 G4GeometryCellStep provides information about the previous and
current "cell" of the particle track.
A "scorer" class derives from the interface G4VScorer. Users may create
customized "scorers" or use the standard scoring.
More on Kernel - M.Asai (SLAC)
57
Plans of event biasing in Geant4




Cross-section biasing for physics processes
General geometrical weight field
 In continuous process for geometrical, angular, energy biasing and
weight window.
 First implementation of weight-window biasing option was
introduced with Geant4 6.0.
Another biasing options are under study.
Other scoring options rather than surface flux counting which is
currently supported are under study.
User’s contribution is welcome.
More on Kernel - M.Asai (SLAC)
58
Fast simulation
(shower parameterization)
Fast simulation - Generalities

Fast Simulation, also called as shower parameterization, is a shortcut to the
"ordinary" tracking.

Fast Simulation allows you to take over the tracking and implement your own
"fast" physics and detector response.

The classical use case of fast simulation is the shower parameterization where
the typical several thousand steps per GeV computed by the tracking are
replaced by a few ten of energy deposits per GeV.

Parameterizations are generally experiment dependent. Geant4 provides a
convenient framework.
More on Kernel - M.Asai (SLAC)
60
Parameterization features

Parameterizations take place in
an envelope. This is typically a
m
mother volume of a sub-system
or of a major module of such a
sub-system.

Parameterizations are often
dependent to particle types
and/or may be applied only to
e
some kinds of particles.

They are often not applied in
complicated regions.
More on Kernel - M.Asai (SLAC)
61
Models and envelope

Concrete models are bound to the envelope
through a G4FastSimulationManager object.

This allows several models to be bound to one
envelope.

« envelope »
(G4LogicalVolume)
The envelope is simply a G4LogicalVolume which
has G4FastSimulationManager.

All its [grand[…]]daughters will be sensitive to the
parameterizations.

A model may returns back to the "ordinary"
tracking the new state of G4Track after
G4FastSimulationManager
ModelForElectrons
ModelForPions
parameterization (alive/killed, new position, new
momentum, etc.) and eventually adds secondaries
(e.g. punch-through) created by the
parameterization.
More on Kernel - M.Asai (SLAC)
62
Fast Simulation

The Fast Simulation components are
indicated in white.
« envelope »
(G4LogicalVolume)
G4FastSimulationManager
Placements

G4Track
G4ProcessManager

Process xxx
Multiple Scattering

G4FastSimulationManagerProcess
G4Transportation

ModelForElectrons
ModelForPions
When the G4Track comes in an envelope,
the G4FastSimulationManagerProcess
looks for a G4FastSimulationManager.
If one exists, at the beginning of each step
in the envelope, each model is asked for a
trigger.
In case a trigger is issued, the model is
applied at the point the G4track is.
Otherwise, the tracking proceeds with a
normal tracking.
More on Kernel - M.Asai (SLAC)
63
G4FastSimulationManagerProcess

The G4FastSimulationManagerProcess is a process providing the interface
between the tracking and the fast simulation.

It has to be set to the particles to be parameterized:

The process ordering must be the following:
[n-3] …
[n-2] Multiple Scattering
[n-1] G4FastSimulationManagerProcess
[ n ] G4Transportation

It can be set as a discrete process or it must be set as a continuous &
discrete process if using ghost volumes.
More on Kernel - M.Asai (SLAC)
64
Ghost Volume

Ghost volumes allow to define envelopes independent to the volumes of the
tracking geometry.

For example, this allows to group together electromagnetic and hadronic
calorimeters for hadron parameterization or to define envelopes for imported
geometries which do not have a hierarchical structure.

In addition, Ghost volumes can be sensitive to particle type, allowing to define
envelops individually to particle types.

Ghost Volume of a given particle type is placed as a clone of the world volume
for tracking.


This is done automatically by G4GlobalFastSimulationManager.
The G4FastSimulationManagerProcess provides the additional navigation inside a
ghost geometry. This special navigation is done transparently to the user.
More on Kernel - M.Asai (SLAC)
65
User support processes
in Geant4
User Support

Geant4 Collaboration offers extensive user supports.

Users workshops

Tutorial courses

HyperNews and mailing list

Bug reporting system

Requirements tracking system

Daily “private” communications

Technical Forum
http://cern.ch/geant4/
More on Kernel - M.Asai (SLAC)
67
Geant4 users workshop

Users workshops were held or are going to be held hosted by several
institutes for various user communities.

KEK - Dec.2000, Jul.2001, Mar.2002, Jul.2002, Mar.2003, Jul.2003,
Jul.2004, Jan.2005

SLAC - Feb.2002

Spain (supported by INFN) - Jul.2002

CERN - Nov.2002

NASA/ESA/Vanderbilt - Jan.2003, May.2004, Oct.2005

Helsinki - Oct.2003, Jun.2005

Local workshops of one or two days were held or are planned at
several places.
More on Kernel - M.Asai (SLAC)
68
Geant4 tutorials / lectures



In addition to the users workshops, many tutorial courses and lectures
with some discussion time slots were held for various user communities.
 CERN School of Computing
 Italian National School for HEP/Nuclear Physicists
 MC2000
 MCNEG workshop
 IEEE NSS/MIC
 KEK, SLAC, DESY, FNAL, INFN, Frascati, Karolinska, GranSasso, etc.
 ATLAS, CMS, LHCb
 Tutorials/lectures at universities
 Italy - Genoa, Bologna, Udine, Roma, Trieste
 U.K. - Imperial
 U.S. - Vanderbilt
Geant4 collaboration is happy to offer tutorial courses if requested.
SLAC Geant4 team is offering tutorial courses regularly.
 http://geant4.slac.stanford.edu/
More on Kernel - M.Asai (SLAC)
69
HyperNews

HyperNews system was set up in April 2001
More on Kernel - M.Asai (SLAC)
70
HyperNews


22 categories
Not only “user-developer”,
but also “user-user”
information exchanges are
quite intensive.
More on Kernel - M.Asai (SLAC)
71
HyperNews is quite active
More on Kernel - M.Asai (SLAC)
72
Some postings are novice…
More on Kernel - M.Asai (SLAC)
73
Some are excellent users contribution
More on Kernel - M.Asai (SLAC)
74
Technical Forum



In the Technical Forum, the Geant4 Collaboration, its user community and
resource providers discuss:
 major user and developer requirements, user and developer priorities,
software implementation issues, prioritized plans, physics validation issues,
user support issues
The Technical Forum is open to all interested parties
 To be held at least 4 times per year (in at least two locales)
 First Technical Forum at TRIUMF in September 2003, followed by forum
meetings at various places.
The purpose of the forum is to:
 Achieve, as much as possible, a mutual understanding of the needs and
plans of users and developers.
 Provide the Geant4 Collaboration with the clearest possible understanding
of the needs of its users.
 Promote the exchange of information about physics validation performed by
Geant4 Collaborators and Geant4 users.
 Promote the exchange of information about user support provided by
Geant4 Collaborators and Geant4 user communities.
More on Kernel - M.Asai (SLAC)
75
SLAC Geant4 team
Scientific Computing and
SLAC Computing Services (SCS)
Computing Services (SCCS)
More on Kernel - M.Asai (SLAC)
77
Physics Experiment Support Group

Objectives

Support software developments
and maintenances for the
experiments SLAC is hosting

Participate nation-wide and
international collaborating
research and development
activities of HEP software

On-line team

Database team

Off-line team
More on Kernel - M.Asai (SLAC)
78
Vis & Analysis Tools for HEP
• Experiment Independent
• Abstract Interfaces
• Component Architecture
BaBar, GLAST,
LCD, Geant4
Abstract Interfaces
IceCube, CLEO
http://www.freehep.org
• Experiment or collaboration choose the
interfaces; end users choose the desktop tool
• Based on well-defined, flexible, language
neutral interfaces
Data Analysis
& Visualization
Event Display
AIDA
HepRep
JAS
Mark Dönszelmann
Tony Johnson
Joseph Perl
Victor Serbo
Max Turri
Open
Scientist
Lizard
WIRED
FRED
More on Kernel - M.Asai (SLAC)
79
Geant4
More on Kernel - M.Asai (SLAC)
80
http://geant4.slac.stanford.edu/
More on Kernel - M.Asai (SLAC)
81
More on Kernel - M.Asai (SLAC)
82