Gabriele Cosmo
151 lines
6.3 KiB
Plaintext
151 lines
6.3 KiB
Plaintext
|
|
Examples for event biasing
|
|
--------------------------
|
|
|
|
This directory includes example applications to demonstrate the usage of
|
|
different variance reduction techniques supported in Geant4, or possible
|
|
from the user applications.
|
|
|
|
|
|
General remark to variance reduction
|
|
------------------------------------
|
|
The tools provided for importance sampling (or geometrical splitting and
|
|
Russian roulette) and for the weight window technique require the user to
|
|
have a good understanding of the physics in the problem. This is because
|
|
the user has to decide which particle types have to be biased, define the
|
|
cells (physical volumes, replicas) and assign importances or weight
|
|
windows to that cells. If this is not done properly it can not be
|
|
expected that the results describe a real experiment. The examples given
|
|
here only demonstrate how to use the tools technically. They don't intend
|
|
to produce physical correct results.
|
|
|
|
General remark to scoring
|
|
-------------------------
|
|
A interface G4VScorer is provided for the user. The user may create his
|
|
own class to perform the desired scoring. The user defined class
|
|
therefore should inherit from the interface G4VScorer.
|
|
An example of an implementation of a scorer is G4Scorer
|
|
which may be found in source/event.
|
|
The scoring in these examples is done with a G4Scorer.
|
|
The variance reduction techniques and scoring do not support all options
|
|
of the Geant4 geometry. It only supports physical volumes and simple
|
|
replicas.
|
|
To identify a physical volume (or replica) objects of the class
|
|
G4GeometryCell are used. Scoring is done according to these
|
|
cells and importance values or the weight windows may be assigned to
|
|
them.
|
|
When scoring is done in a parallel geometry special action has to be taken
|
|
to prevent counting of "collisions" with boundaries of the mass geometry
|
|
as interactions. This is differently handled when scoring is done in the
|
|
mass geometry.
|
|
|
|
--> G4GeometryCell of the parallel geometry must not share boundaries with
|
|
the world volume! <--
|
|
|
|
Known problems
|
|
--------------
|
|
In the following scenario it can happen that a particle is not
|
|
biased and it's weight is therefore not changed even if it crosses
|
|
a boundary where biasing should happen.
|
|
Importance and weight window sampling create particles on boundaries
|
|
between volumes. If the GPIL method of a physical process returns
|
|
0 as step length for a particle on a boundary and if the PostStepDoIt of
|
|
that process changes the direction of the particle to go back in the
|
|
former volume the biasing won't be invoked.
|
|
This will produce particles with weights that do not correspondent to the
|
|
importance of the current volumes.
|
|
|
|
Further information:
|
|
--------------------
|
|
Short description of importance sampling and scoring:
|
|
http://cern.ch/geant4/working_groups/geometry/biasing/Sampling.html
|
|
|
|
Example B01
|
|
===========
|
|
|
|
The example uses importance sampling or the weight window technique
|
|
according to an input parameter. It uses scoring in both cases.
|
|
Importance values or weight windows are defined according to the mass
|
|
geometry. In this example the weight window technique is configured such
|
|
that it behaves equivalent to importance sampling: The window is actually
|
|
not a window but simply the inverse of the importance value and only
|
|
one energy region is used that covers all energies in the problem.
|
|
The user may change the weight window configuration by changing the
|
|
initialization of the weight window algorithm in example,cc.
|
|
Different energy bounds for the weight window technique may be specified
|
|
in B01DetectorConstruction.
|
|
|
|
The executable takes one optional argument: 0 or 1. Without argument or
|
|
with argument: 0, the importance sampling is applied with argument: 1,
|
|
the weight window technique is applied.
|
|
|
|
|
|
Example B02
|
|
===========
|
|
|
|
This example uses a parallel geometry to define G4GeometryCell objects
|
|
for scoring and importance sampling. In addition it customizes
|
|
the scoring. In this example one scorer creates a histogram.
|
|
|
|
Compiling and running
|
|
---------------------
|
|
Can be compiled and executed on a RedHat-7.3 system with gcc-3.2.3
|
|
compiler and the tcsh shell.
|
|
To compile this example you need AIDA 3.2.1 installed. To link
|
|
and run it you need a AIDA compliant analysis package. The
|
|
GNUmakefile of this example shows how to use AIDA through PI as
|
|
analysis interface.
|
|
Histograms are saved in HBOOK format. It can be displayed with PAW or
|
|
compatible packages.
|
|
You need to set the following variables in your environment:
|
|
"G4ANALYSIS_USE"
|
|
"PI_BASE_DIR" (where PI has been installed)
|
|
Finally, source the script setupPI.csh.
|
|
|
|
Now you should be able to run gmake and to run exampleB02.
|
|
|
|
The example stores the plot in the file b02.hbook.
|
|
To look at the histogram using lizard you also may use PI 1.2.1
|
|
http://cern.ch/PI.
|
|
|
|
|
|
Example B03
|
|
===========
|
|
|
|
This example uses Geant4 and in particular importance sampling and
|
|
scoring through python. It creates a simple histogram. It's meant
|
|
to demonstrate how to use a customized scorer and importance sampling
|
|
in combination with a scripting language, python.
|
|
Geant4 code is executed from a python session. Therefore, swig 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 PI: http://cern.ch/PI.
|
|
At the end a histogram called "trackentering.hbook" is created and can be
|
|
displayed using standard packages (such as PAW).
|
|
|
|
Building, compiling and running
|
|
-------------------------------
|
|
|
|
You need to set the following variables in your environment:
|
|
"G4ANALYSIS_USE"
|
|
"PI_BASE_DIR" (where PI has been installed)
|
|
"SWIG_BASE_DIR" (where SWIG 1.3.15 has been installed)
|
|
Finally source the script setupPI.csh.
|
|
|
|
You may run gmake now.
|
|
You should be able to execute the file B03RunApplication.py from your
|
|
shell or from a lizard session now.
|
|
|
|
At the moment the plotting is not available using a python script, but
|
|
it is planned in future releases. A histogram is created and later displayed
|
|
using standard analysis packages.
|
|
|
|
To clean all the added files, just type gmake clean_all.
|
|
|
|
Files in B03;
|
|
B03Application.py: Is a example class utilizing importance sampling
|
|
and scoring using python.
|
|
B03RunApplication.py: Is a python script running the example.
|
|
It may be executed from the shell or in a python session.
|
|
B03App.py: Is created by swig using swig.
|