Gabriele Cosmo
277 lines
12 KiB
Plaintext
277 lines
12 KiB
Plaintext
-------------------------------------------------------------------
|
|
|
|
=========================================================
|
|
Geant4 - an Object-Oriented Toolkit for Simulation in HEP
|
|
=========================================================
|
|
|
|
field01
|
|
-------
|
|
|
|
Example that enables investigation of the accuracy and performance of the
|
|
tracking in a magnetic field.
|
|
|
|
The key Geant4 capabilities demonstrated in this example are:
|
|
- creating a uniform magnetic field interactively using the field
|
|
messenger,
|
|
- choosing the type of Runge Kutta stepper used for integration of the
|
|
motion of charged particles in the magnetic field,
|
|
- controlling the thresholds that determine which looping particles are
|
|
killed by G4Transporation.
|
|
|
|
Some of these capabilities are available via interactive commands,
|
|
implemented in F01FieldMessenger.
|
|
|
|
A. The magnetic field is defined in F01FieldSetup, which is created in
|
|
the ConstructSDandField() method in the F01DetectorConstruction
|
|
class.
|
|
|
|
B. Choosing the type of stepper -
|
|
|
|
The basic capabilities of choosing the stepper type are demonstrated in the
|
|
field.in macro file:
|
|
|
|
/field/setStepperType 145 ## Choose a stepper type ( Tsitouras )
|
|
|
|
/field/setStepperType 101 ## Choose an FSAL stepper ( FEqRK1 )
|
|
|
|
/field/setMinStep 0.1 mm ## Smaller steps always succeed
|
|
|
|
/field/update ## Initialise using parameters above
|
|
|
|
In addition it is possible to choose to use a new type of stepper, known
|
|
as 'First Same as Last' or FSAL, which in each step obtains the field value
|
|
at the step endpoint and evaluates the 'right hand size' of the equation
|
|
for the next integration step. This reduces the number of calls to the field
|
|
evaluation, which can be one the most computationally expensive methods,
|
|
while providing similar accuracy.
|
|
|
|
There are several potential choices of the stepper type. Here are some
|
|
suggestions:
|
|
===========================================================================
|
|
Number Name of Stepper Comments
|
|
===========================================================================
|
|
Recommended - default since Geant4 10.4:
|
|
|
|
15 - 'DoPri5' or
|
|
Dormand Prince 745 : Uses a pair 4th & 5th order formulae (like other 4/5
|
|
well-known and very efficient embedded method
|
|
methods); their difference is the error estimate.
|
|
Highly recommended in literature, including
|
|
Hairer & Wanner, & Numerical Recipes
|
|
Used in several established RK code (e.g. DOPRI5)
|
|
===========================================================================
|
|
Good choices for reasonably smooth fields:
|
|
|
|
45 - BogackiShampine45 : more efficient embedded 4/5 pair
|
|
Used in many applications, including
|
|
RKSUITE suite.
|
|
|
|
145 - Tsitouras45 : potentially the most efficient embedded 4/5
|
|
pair - found in expanded search of parameter
|
|
space.
|
|
|
|
56 - Dormand Prince RK56 : higher order embedded method from authors of DoPri5.
|
|
Uses a pair 5th & 6th order formulae.
|
|
|
|
78 - Dormand Prince RK78 : higher order embedded method from authors of DoPri5.
|
|
Uses a pair 7th & 8th order formulae.
|
|
|
|
9 - NystromRK4 : a specialised Nystrom method for magnetic fields.
|
|
Reuses the field value at the mid-point of the step,
|
|
and also provides an analytical estimation of the
|
|
integration error based on numerical evaluation of
|
|
fourth order variation in the equation for
|
|
magnetic field.
|
|
===========================================================================
|
|
The new 'First Same as Last' (FSAL) steppers can be chosen in addition:
|
|
1 - RKFEq1 : FSAL stepper with improved equilibrium properties.
|
|
When kinks or other anomalies are encountered,
|
|
and at the start of integration when the best
|
|
step size is not known, this type of stepper
|
|
converges faster and more smoothly to good
|
|
step sizes.
|
|
===========================================================================
|
|
The old default and old first alternative -
|
|
|
|
4 - ClassicalRK4 : original Runge-Kutta method, very robust but slower )
|
|
( obtains error estimate by doing 2 half steps )
|
|
Good baseline for comparison - long experience of use.
|
|
May be good alternative for less smooth fields.
|
|
|
|
8 - Cash Karp RKF 45 : The oldest 'embedded' RK method in Geant4 -
|
|
also fairly robust.
|
|
Faster than ClassicalRK4 for smoother fields,
|
|
as it does not need two half steps to estimate error.
|
|
Available since Geant4 1.0
|
|
|
|
===========================================================================
|
|
Other potential choices for non-smooth fields (with kinks, abrupt changes):
|
|
|
|
3 - SimpleHeum : low order, with error obtained from half-steps
|
|
23 - BogackiShampine23 : lower order embedded method (new in 10.3-beta)
|
|
===========================================================================
|
|
|
|
|
|
|
|
C. Controlling the killing of looping particles
|
|
|
|
|
|
Occasionally tracks 'looping' in a strong magnetic field, making little
|
|
progress even over hundreds of integration steps. This is due to a
|
|
combination of a strong magnetic field and a thin material (gas or vacuum)
|
|
in which the size of a physics step is substantially larger than the radius
|
|
of curvature of the track.
|
|
|
|
Since the amount of CPU time which can be consumed by one or few such tracks
|
|
is very large, it is important to limit the number of integration steps
|
|
spent on these tracks. The module for propagation in field in Geant4
|
|
flags tracks which take more than a certain number (default 1,000) integration
|
|
steps without reaching the requested end of the step size, which was
|
|
determined by the physics and geometry.
|
|
|
|
The Geant4 G4Transportation and G4CoupledTransportation processes are tasked
|
|
to select which of the looping tracks are killed and which survive. To
|
|
balance the potential significant cost of integrating looping particles,
|
|
three thresholds exist
|
|
|
|
The 'Warning' Energy: a track with energy below this value that is found to
|
|
loop is killed silently (no warning.)
|
|
|
|
Above the 'Warning Energy', if a track is selected for killing a warning is
|
|
generated.
|
|
|
|
The 'Important' Energy: the threshold energy above which a track will survive
|
|
for multiple steps if found looping.
|
|
|
|
number of 'tracking' steps. They will be only be killed only if they still
|
|
loop after than
|
|
The number of 'trials': the number of steps that 'important' tracks survive.
|
|
|
|
Note that currently only stable particles are killed. ( Refinements to enable
|
|
toggling whether unstable particles can be killed are in development. )
|
|
|
|
This example demonstrate choosing different values for these parametes
|
|
in the main() method of field01.cc using one of two techniques.
|
|
|
|
The first method is new in Geant4 release 10.5, and uses the G4PhysicsListHelper
|
|
which has methods to choose a pre-selected set of parameter values. The choices
|
|
are between a set each of low and high thresholds. Either one can be enabled
|
|
by calling correspondingly
|
|
- G4PhysicsListHelper::GetPhysicsListHelper()->UseLowLooperThresholds();
|
|
or
|
|
- G4PhysicsListHelper::GetPhysicsListHelper()->UseHighLooperThresholds();
|
|
These methods must be called before the physics is constructed - i.e. typically
|
|
before RunManager's Initialise() method is called.
|
|
This works only if either
|
|
- a modular physics lists is used, or if
|
|
- the G4ModularPhysicsList and its AddTransporation method are used
|
|
to create and register a common transportation process for all particles
|
|
(one for each thread).
|
|
|
|
ii) Fine grained control (available in Geant4 versions since 7.0)
|
|
|
|
Fine grained control of the Transportation's parameters for looping particles
|
|
is also possible.
|
|
|
|
This is demonstrated in the F01RunAction's ChangeLooperParameters method,
|
|
which is called by the BeginOfRunAction. There the appropriate
|
|
Transportation object for the electron is obtained, and its parameters
|
|
(if valid) are used to overwrite the thresholds in the G4Transportation class.
|
|
|
|
For example, to ensure that only looping particles with energy 10 keV are
|
|
killed silently we change the value of the 'Warning' Energy:
|
|
|
|
runAction->SetWarningEnergy( 10.0 * CLHEP::keV );
|
|
|
|
[ This is passed along to the registered G4Transportation or
|
|
G4CoupledTransportation object by the F01RunAction's ChangeLooperParameters.]
|
|
|
|
As a result the killing of any (stable) looping track with energy over 10 keV
|
|
will generate a warning.
|
|
|
|
A second configurable energy threshold enables tracks above it to survive a
|
|
chosen number of 'tracking' steps. They will be only be killed only if they
|
|
still loop after than number of tracking steps. F01RunAction's methods are
|
|
used to configure these parameters:
|
|
|
|
runAction->SetImportantEnergy( 0.1 * CLHEP::MeV );
|
|
runAction->SetNumberOfTrials( 30 );
|
|
|
|
which the run action passes to the G4Transportation or
|
|
G4CoupledTransportation object registered for the electron.
|
|
|
|
Note that for all pre-configured and modular physics lists share a single
|
|
Transportation process for all types of particles. So the parameters for
|
|
killing loopers will be shared by all particle types in this case.
|
|
|
|
|
|
Background Information
|
|
|
|
1- GEOMETRY DEFINITION
|
|
|
|
The "Absorber" is a solid made of a given material.
|
|
|
|
Three parameters define the absorber :
|
|
- the material of the absorber,
|
|
- the thickness of an absorber,
|
|
- the transverse size of the absorber (the input face is a square).
|
|
|
|
The volume "World" contains the "Absorber".
|
|
In this test the parameters of the "World" can be changed , too.
|
|
|
|
In addition a transverse uniform magnetic field can be applied.
|
|
|
|
The default geometry is constructed in F01DetectorConstruction class,
|
|
but all the parameters can be changed via
|
|
the commands defined in the F01DetectorMessenger class.
|
|
|
|
2- AN EVENT : THE PRIMARY GENERATOR
|
|
|
|
The primary kinematic consists of a single particle (electron, Ekin = 0.5 GeV)
|
|
which hits the
|
|
absorber perpendicular to the input face. The type of the particle
|
|
and its energy are set in the F01PrimaryGeneratorAction class, and can
|
|
be changed via the G4 build-in commands of G4ParticleGun class (see
|
|
the macros provided with this example).
|
|
|
|
It is also possible to change the position of the primary particle vertex
|
|
or activate its randomization via the commands defined in the
|
|
F01PrimaryGeneratorMessenger class.
|
|
|
|
A RUN is a set of events.
|
|
|
|
3- DETECTOR RESPONSE
|
|
|
|
The spatial distribution of charged particles transported in magnetic
|
|
field is envistigated.
|
|
A HIT is a record, event per event , of all the
|
|
informations needed to simulate and analyse the detector response.
|
|
|
|
In this example a F01CalorHit is defined as a set of 2 informations:
|
|
- the total energy deposit in the absorber,
|
|
- the total tracklength of all charged particles in the absorber,
|
|
|
|
Therefore the absorber is declared
|
|
'sensitive detector' (SD), see F01CalorimeterSD, which means they can contribute to the hit.
|
|
|
|
4- PHYSICS LIST
|
|
|
|
The particle's type and the physic processes which will be available
|
|
in this example are set in the FTFP_BERT physics list. This physics list
|
|
requires data files for electromagnetic and hadronic processes.
|
|
See more on installation of the datasets in Geant4 Installation Guide,
|
|
|
|
5- HOW TO START ?
|
|
|
|
- Execute field01 in 'batch' mode from macro file e.g.
|
|
% ./field01 field01.in
|
|
|
|
- Execute field01 in 'interactive' mode with visualization e.g.
|
|
% ./field01
|
|
....
|
|
Idle> /run/beamOn 1
|
|
....
|
|
|
|
|
|
|