Re: unphysical spectra

From: Alfredo Ferrari (alfredo.ferrari@cern.ch)
Date: Thu Oct 23 2003 - 17:09:05 CEST

  • Next message: John Clem: "Fw: fluka"

    Hi Ludwig

    I would like to correct your statements

    > all spectra from the neutron transport
    > at low energies are expected to show unphysical shape

    of course, all particle spectra are expected to show the physically
    correct shape. What you cannot do in FLUKA or in whichever other code
    based on evaluated nuclear file libraries is to expect correlation at
    SINGLE scattering level, which is very different from particle spectra
    which of course should be ok.
    On top of this energy deposition, again as in all these
    code, is based on kerma factors (sometimes called heating numbers for
    reactor like applications). So you CANNOT get the energy deposition
    spectra, because the code is actually depositing always the same, average,
    amount of energy for each given neutron energy.

    As I tried to explain this is an (unfortunate) consequence of both how
    nuclear data files are organized and multigroup cross sections are built
    (similar problems exist for pointwise cross sections as well).
    At 19.7 MeV you (by .1 MeV) are still running in the "model" part of the
    code, and therefore, at least until the first scattering, things are
    exclusively reproduced. Below 19.6 the nuclear data treatment comes in
    with all its pros (very precise, data driven, no other chance to treat
    accurately resonances etc) and cons (almost impossible to make "exclusive"
    interactions, which sometimes, and your problem is an example, are of
    interest). As I told similar limitations are intrinsic to all neutron
    transport codes in the energy range where they are based on libraries.

    If you need to compute detector efficiencies, you have two choices:

    a) simple one: your detector is "simple" and the cross section for
       the reaction of interest well known and "hits" are just "counted", ie a
       3-He or BF3 counter possibly inside moderator/attenuator assemblies.
       In this case all neutron codes can give you good estimates of neutron
       spectra inside the sensitive volume, that you can easily fold with
       the reaction of interest cross section and compute reliably
       efficiencies (ie we did it many times for REM counters, Bonner spheres
       etc)
    b) you really need to compute the energy deposition spectra because the
       detector you use is really seeing them. Hard task, we did it twice
       and in both cases we had to do ourselves our own "evaluation" of
       the relevant isotopes cross sections in exclusive mode (with quite some
       invention sometimes). The typical example is a liquid scintillator
       where you want recoil spectra. On top of these there is also to take
       into account that often "heavy" recoils do not ionize or produce
       much less light per unit deposited energy of light ones and therefore
       not only the energy deposition spectra, but the dE/dx of the recoiling
       ions are an issue in determining the detector response. The example
       you mention seems to fall into this category, you need the recoil
       spectra and you need the particle ids of these recoils (almost all
       crystals emit different light outputs in response to 10 keV energy
       deposition by electrons, rather than the same amount deposited by 1 MeV
       protons, or some heavy recoiling nucleus, the latter being often
       almost undetectable).
       I have no easy recipe to treat this problem, in principle you would
       need to make "exclusive" all neutron interactions below 19.6 MeV for
       all isotopes in your crystal composition. This can be easily done
       for elastic scattering (and I can help on this), already much more
       complex on (n,n') and (n,gamma) (because you need photon deexcitation
       cascades correlated, while they are recorded inclusively in the nuclear
       data files), very very hard for (n,2n), (n,na) etc. We did this for
       Argon (we are in a neutrino experiments with thousands tons of liquid
       Argon), for 6-Li, for hydrogen (easy of course), for Cd and Xe limited
       to (n,gamma) and we are doing for Carbon. It was already quite a lot
       of work.

                          Ciao
                         Alfredo

    On Wed, 22 Oct 2003, Ludwik Pienkowski wrote:

    > Dear FLUKA authors,
    >
    > Thanks to Alfredo for the explanation concerning the strange cases
    > that show single neutron interactions at low energy. If I well
    > understand this explanation, all spectra from the neutron transport
    > at low energies are expected to show unphysical shape, and only
    > the average values of the variables are to be considered meaningful.
    >
    > Please look at the following test in which neutrons of two close
    > energies 19.5 MeV and 19.7 MeV hit a virtually infinitely thick
    > lead target. The program (attached to this email) counts the total
    > energy deposition using the routine mgdraw. The energy deposition
    > spectra are shown in file n2test.pdf. It is observed that only
    > average energy depositions are similar for both energies
    > (19.5MeV and 19.7MeV), while all other observables differ wildly.
    > And so:
    > 1.RMS=22.7 and 9.9 MeV, for the two energies, respectively
    > 2.a very strange shape at 19.5 MeV and a nice gaussian shape
    > at 19.7 MeV
    >
    > Moreover both spectra show a long tail. There are many events
    > that show energy deposition much larger than the available energy.
    >
    > This test run is closely related to my main task for which I'm
    > trying to use FLUKA. The task is to estimate the neutron detection
    > probability inside a detector, for example inside a solid crystal
    > cube of a volume of about 1dcm3. This probability I'm trying to
    > estimate from the energy deposition spectra looking for the events
    > that deposit energy inside the crystal above a given threshold.
    > However, the observations from the presented test run show me that
    > I'm not sure that I can use FLUKA for this purpose.
    >
    > Would you have any suggestions for me on whether and how could
    > I use FLUKA to treat the above problem.
    >
    > Best regards,
    > Ludwik Pienkowski
    >
    >
    > P.S.The attached files input files:
    >
    > 1.mgdraw-n2test.f
    > 2.lead.peg
    > 3.n2test195.inp
    > 4.n2test197.inp
    > and energy deposinon spectra figure:
    > 5.n2test.pdf
    >
    > link command:
    > lfluka -o n2test -m fluka mgdraw-n2test.f
    >
    > fluka run:
    > rfluka -e n2test -N 0 -M 1 -p lead n2test195
    > rfluka -e n2test -N 0 -M 1 -p lead n2test197
    >
    > The results are in the files: n2test195001.log, n2test197001.log
    > and n2test.pdf shows the graphs.
    >
    >
    > ================================
    > Heavy Ion Laboratory
    > Warsaw University
    > ul. Pasteura 5A
    > 02-093 Warszawa
    > Poland
    > pienkows@slcj.uw.edu.pl
    > ================================
    >
    >

    -- 
    +----------------------------------------------------------------------------+
    |  Alfredo Ferrari                ||  Tel.: +41.22.767.6119                  |
    |  CERN-AB                        ||  Fax.: +41.22.767.7555                  |
    |  1211 Geneva 23                 ||  e-mail: Alfredo.Ferrari@cern.ch        |
    |  Switzerland                    ||          Alfredo.Ferrari@mi.infn.it     |
    +----------------------------------------------------------------------------+
    

  • Next message: John Clem: "Fw: fluka"

    This archive was generated by hypermail 2.1.6 : Thu Oct 23 2003 - 17:09:23 CEST