Warning, /geant4/examples/extended/biasing/GB03/README is written in an unsupported language. File is not indexed.
0001 Example GB03 : geometry based biasing
0002 -------------------------------------
0003
0004 This example illustrates a use of generic biasing classes to implement a
0005 technique near to "geometry importance biasing".
0006
0007 The geometry is the same than in EM tests, with the sampling calorimeter
0008 made of a series of layers of absorber and gap.
0009
0010 The biasing applies to neutrons only.
0011
0012 Instead of explicitely assigning "importance" values to the layers, we
0013 split neutrons moving forward and kill the ones moving backward, when they
0014 reach the exit of an absorber volume.
0015
0016 The splitting factor can be controlled by command line, eg:
0017
0018 /GB03/biasing/setSplittingFactor 2
0019
0020 which also determines the killing probability : 1/(splitting factor).
0021
0022 It can be seen than when defining 10 layers (see exampleGB03.in), a
0023 splitting factor 2 works fine : we don't suffer from under- or over-splitting.
0024 If going to 20 layers, then a splitting with a factor 2 is too large,
0025 and the biasing suffers from over-splitting. (And we can not go lower than
0026 "2", which would mean "1" and hence, no biasing...)
0027
0028 To alleviate the over-splitting, we introduce a probability to apply the
0029 splitting (and killing) (this is one solution, others can be considered), that
0030 can be changed as:
0031
0032 /GB03/biasing/setApplyProbability 0.5
0033
0034 With above value, we can see that we recover a satisfactory biasing scheme,
0035 with neutrons penetrating the entire setup, without over-splitting.
0036
0037
0038 The classes involved are:
0039
0040 - GB03BOptnSplitOrKillOnBoundary : which is the biasing operation making
0041 the splitting and killing;
0042 - GB03BOptrGeometryBasedBiasing : which is the biasing operator, making
0043 decision to use above operation, and configuring it, passing it the
0044 splitting factor and probability to apply the biasing.
0045
