FLUKA: WW-FACTOr
Previous Index Next
WW-FACTOr
defines Weight Windows in selected regions
See also BIASING, WW-THRESh, WW-PROFIle
Attention: Option WW-FACTOr alone is not sufficient to define a
weight window. One or more WW-THRESH cards are also necessary in
order to activate the window.
WHAT(1)
>= 0.0 : Russian Roulette (RR) parameter (Window "bottom" weight
at the lower energy threshold set by WW-THRESh).
< 0.0 : resets to -1.0 a possible positive value set in a
previous WW-FACTOr card
This value can be modified by WHAT(4)
in option WW-THRESh or
by WHAT(2)
in WW-PROFIle, and can be overridden in the user
routine UBSSET (argument WWLOW in the calling list, see 13})
Default
= -1.0 (no RR)
WHAT(2)
> 1.7*WHAT(1)
: Splitting parameter (Window "top" weight at
the lower energy threshold set by WW-THRESh)
= 0.0 : ignored
=< 1.7*WHAT(1)
: resets to infinity (no splitting) a possible
value set in a previous WW-FACTOr card
This value can be modified by WHAT(4)
in option WW-THRESh or
by WHAT(2)
in WW-PROFIle, and can be overridden in the user
routine UBSSET (argument WWHIG in the calling list, see 13})
Default
: infinity (no splitting)
WHAT(3)
> 0.0 : Multiplicative factor to be applied to the two
energy thresholds for RR/splitting (defined by
option WW-THRESh) in the region of interest
= 0.0 : ignored
< 0.0 : resets to 1.0 a possible value set in a previous
WW-FACTOr card
This value can be overridden in the user routine UBSSET
(argument WWMUL in the calling list, see 13})
Default
: 1.0 (RR/splitting thresholds are not modified)
WHAT(4)
= lower bound of the region indices (or corresponding name) in
which the indicated RR and/or splitting parameters apply
("From region WHAT(4)
...")
Default
= 2.0
WHAT(5)
= upper bound of the region indices (or corresponding name) in
which the indicated RR and/or splitting parameters apply
("...to region WHAT(5)
...")
Default
= WHAT(4)
WHAT(6)
= step length in assigning indices. ("...in steps of
WHAT(6)
").
Default
= 1.0
SDUM
: a number from 1.0 to 5.0 in any position, indicating the
low-energy neutron weight-window profile to be applied
in the regions selected (see WW-PROFIle). Exceptionally,
here SDUM
must be a number, in free format, rather than
a character string.
= blank, zero or non numerical: ignored
< 0.0 : resets to 1.0 a possible value previously given
This value can be overridden in the user routine UBSSET
(argument JWSHPP in the calling list, see 13})
Default
(if no WW-PROFIle card is present): profile number 1
Default
(option WW-FACTOr or WW-THRESH not given): no weight window
Notes:
1) This option, which must be used together with WW-THRESh, allows
the user to define a very detailed weight-window for Russian
Roulette and splitting: energy-dependent, per region and per
particle.
WW-THRESh is used to set two basic energy values for each
particle (including electrons and photons but not low-energy
neutrons). From each basic couple of energies, a different
couple of thresholds is generated for each region by
multiplication with the factor provided in WHAT(3)
. A weight
window of minimum width is defined at the lower threshold by
its bottom and top edges (WHAT(1)
and WHAT(2)
); a second wider
window is obtained from it at the higher threshold by
increasing the "top edge" (splitting level) and decreasing the
"bottom edge" (RR level) by the amplification factor given
with WW-THRESh. The whole energy range is thus divided in
three parts. In the high-energy part (above the higher
threshold) the window is of infinite width, i.e. no
splitting/RR takes place. In the medium-energy range the
window narrows down continuously with decreasing energy, its
top and bottom edges varying linearly with energy between the
two thresholds. In the low-energy range the window width remains
constant and equal to the minimum value it has at the lower threshold.
2) Russian Roulette is played in a given region if the particle
weight is lower than the bottom window edge for that energy,
particle and region. The particle survives with a probability
equal to the ratio between its weight and the RR edge, and is
given a new weight equal to the RR edge itself.
Splitting is performed if the particle weight is higher than
the top window edge for that energy, particle and region. The
particle is replaced by two identical ones with half its
weight. Note that the top edge must always be at least a
factor two higher than the bottom one, in order to avoid
repeated and useless changes of weight. Actually, it is
suggested to never make this factor less than 3 or 4.
3) For low-energy neutrons, a different scheme applies. Instead
of dividing the energy range into three parts (constant
window, continuously varying window, infinite window), the
window is assigned group by group by means of option
WW-PROFIle, creating a so-called "weight-window profile". On
the other hand, it is not possible to assign a different
profile to each region, but only a maximum of 5 different
profiles are allowed.
4) A form of splitting and Russian Roulette is also provided by
option BIASING. The two options, however, are different in
many respects:
- with WW-FACTOr, splitting and RR are played at the moment
a particle is taken from the stack and starts to be
transported. With BIASING, splitting/RR happens when a
particle crosses a boundary (in the case of hadrons also
- on request - before loading in stack the secondaries
from an inelastic hadron collision)
- while the criterion used by BIASING to trigger splitting/RR
depends only on the RELATIVE IMPORTANCE of various regions
of phase space, the weight window is based on ABSOLUTE
WEIGHT STANDARDS pre-assigned to different phase space
regions
- BIASING can have two purposes: when used at collisions with
RR only, i.e. reducing factor < 1, it aims at increasing the
total number of histories simulated in a given time, namely
to sample over a more extended part of phase space (e.g.
more primary interactions) without leaving any important
part not sufficiently represented. (This is also true of
leading particle biasing for electrons and photons via
option EMF-BIAS). At the same time (and this holds also for
splitting, especially when the option is used at boundary
crossing) it can be applied to sample preferentially from
those regions of phase space which contribute more to the
result.
This second purpose is also that of the WW-FACTOr weight
window, but in addition this option has the advantage to
avoid excessive weight fluctuations. These can be dangerous
in two ways. In general, if transport is biased and no
control is kept on particle weight, it can happen that too
much time is wasted by tracking particles of very low weight
which can only contribute little to the score. On the other
hand, too large weights can also be a problem. If the part
of phase space used for scoring (the "detector") is very
small (typically an element of a "binning" mesh), so that only
a limited number of particles have a chance to enter it, it is
statistically important that they all make contributions of
the same order. A rare particle of large weight crossing the
detector would give rise to an anomalous score not
compensated by opposite fluctuations.
Why should one then use BIASING and not the weight window?
The answer is that "tuning" an absolute weight by region and
energy is more powerful but also much more difficult and
time-consuming than just quantifying relative spatial
importances. In general, it requires a lot of experience
which can often be obtained only by performing repeated runs
of the same case and by making a careful statistical
analysis of history distributions in phase space. Not all
problems are worth of it and not all users are able to do it.
It can also be said that WW-FACTOr and BIASING (and other
non-analogue transport options) are not necessarily mutually
exclusive; on the contrary the weight window can be
successfully used to damp excessive fluctuations originated
by other techniques. However, it is the user's
responsibility to ensure that the average absolute weights
produced independently by the different options be of the
same order of magnitude. Otherwise, possible conflicts could
give rise to a waste of time due to an excessive rate of
weight adjustments, and even to incorrect results.
5) The weight limits defined by WW-FACTOr apply to all particles:
however, it is possible to set different values for specific
particles (see WHAT(3)
of option WW-THRESh). This is especially
necessary when secondary particles are generated with a weight
much smaller than the parent particles of a different kind (for
instance, as the result of LAM-BIAS option).
6) WW-FACTOr is one of the two FLUKA options where SDUM
is used to
input numerical data. (Actually, the material number is
first read as a string and then an internal reading is
performed on the string to get the number).
Example 1 (number based):
WW-FACTOR 76.0 1200.0 1.0 5.0 6.0 0.0
The same example, name based:
WW-FACTOR 76.0 1200.0 1.0 Regfive Regsix 0.0
Example 2 (number based):
WW-FACTOR 13.0 120.0 1.5 27.0 31.0 2.0 3.
The same example, name based:
WW-FACTOR 13.0 120.0 1.5 regio27 regio31 2.0 3.
Previous Index Next
