5 Interactions of Beams with Surroundings
189
The generalized evaporation model GEM treats the fragmentation and deexcitation of the spectator nuclei and includes 66 different ejectiles (up to Mg)
as well as fission processes. The production of radioactive nuclides, both from
projectile and target nuclei, thus follows directly from the microscopic interaction
models.
The transport of low energy neutrons employs cross sections from evaluated
nuclear data libraries such as JENDL below 20 MeV. Electromagnetic interactions
are simulated based on the EGS5 code in the energy range between 1 keV and 1 TeV.
Due to its capability to transport nuclei PHITS is frequently applied in iontherapy and space radiation studies. The code is also used for general radiation
transport simulations, such as in the design of spallation neutron sources.
5.1.2.6 Simulation Uncertainties
Depending on the complexity of Monte-Carlo simulations, the size of geometries
and the energy range involved, calculations can carry considerable uncertainties
of various sources which are in many cases difficult to evaluate. While statistical
uncertainties are generally below a few percent thanks to the available computing
power, systematic errors are in most cases very difficult, or even impossible, to
predict in an accurate way.
The main sources of error are:
• Error due to the physics modelling: e.g., in the uncertainty in cross sections,
especially at modern accelerators operating at very high energies or applications
sensitive to other uncertainties in the modelling used in the simulation code. One
can usually expect up to few 10% uncertainty on integral quantities, while for
multi differential quantities the uncertainty can be much worse.
• Further uncertainties due to the assumptions used in the description of the
geometry and of the materials under study. Usually it is difficult to quantify
this uncertainty and experience shows that a factor of 2 can be taken as a safe
limit for general calculations. For special cases, even in case of rather complex
geometries, but when the design is implemented to a very detailed level, the latter
can be reduced to about 10–20%, but rarely significantly below.
• Typical for accelerator applications, additional errors appear when having beams
grazing at small angles to surfaces, where either the surface roughness is not
taken into account, small misalignments can have large effects, or where one is
interested in scattering effects over large distances. Therefore, especially for the
latter, a safety factor of 2–3 has to be considered.
Only a detailed case-by-case analysis and careful evaluation with monitoring data
can reduce the above to very low levels. Such studies are continuously done by code
developers, as well as core code users and nicely show the possible high-accuracy
reached with modern Monte-Carlo codes (see for instance [10]).
189
The generalized evaporation model GEM treats the fragmentation and deexcitation of the spectator nuclei and includes 66 different ejectiles (up to Mg)
as well as fission processes. The production of radioactive nuclides, both from
projectile and target nuclei, thus follows directly from the microscopic interaction
models.
The transport of low energy neutrons employs cross sections from evaluated
nuclear data libraries such as JENDL below 20 MeV. Electromagnetic interactions
are simulated based on the EGS5 code in the energy range between 1 keV and 1 TeV.
Due to its capability to transport nuclei PHITS is frequently applied in iontherapy and space radiation studies. The code is also used for general radiation
transport simulations, such as in the design of spallation neutron sources.
5.1.2.6 Simulation Uncertainties
Depending on the complexity of Monte-Carlo simulations, the size of geometries
and the energy range involved, calculations can carry considerable uncertainties
of various sources which are in many cases difficult to evaluate. While statistical
uncertainties are generally below a few percent thanks to the available computing
power, systematic errors are in most cases very difficult, or even impossible, to
predict in an accurate way.
The main sources of error are:
• Error due to the physics modelling: e.g., in the uncertainty in cross sections,
especially at modern accelerators operating at very high energies or applications
sensitive to other uncertainties in the modelling used in the simulation code. One
can usually expect up to few 10% uncertainty on integral quantities, while for
multi differential quantities the uncertainty can be much worse.
• Further uncertainties due to the assumptions used in the description of the
geometry and of the materials under study. Usually it is difficult to quantify
this uncertainty and experience shows that a factor of 2 can be taken as a safe
limit for general calculations. For special cases, even in case of rather complex
geometries, but when the design is implemented to a very detailed level, the latter
can be reduced to about 10–20%, but rarely significantly below.
• Typical for accelerator applications, additional errors appear when having beams
grazing at small angles to surfaces, where either the surface roughness is not
taken into account, small misalignments can have large effects, or where one is
interested in scattering effects over large distances. Therefore, especially for the
latter, a safety factor of 2–3 has to be considered.
Only a detailed case-by-case analysis and careful evaluation with monitoring data
can reduce the above to very low levels. Such studies are continuously done by code
developers, as well as core code users and nicely show the possible high-accuracy
reached with modern Monte-Carlo codes (see for instance [10]).