Re: [fluka-discuss]: RESNUCL with LAM-BIAS

From: Alberto Fasso' <fasso_at_slac.stanford.edu>
Date: Thu, 26 Feb 2015 06:41:00 -0800 (PST)

Hi Hayley and George,

I don't agree with several statements made by George regarding Hayley's
problem. I will consider them separately.

On Wed, 25 Feb 2015, George Kharashvili wrote:

> Hi Hayley,
>
> I think my earlier comment about overbiasing may not be correct. I see that i$
> one of the examples used during the last advanced course they used -5E-4
> biasing factor for hadronic inelastic interactions.

Overbiasing: a biasing factor of 1.E-6 is very likely to be too much. I don't
have a precise reason for saying that, except that the "golden rule" of biasing
is to do it gently, without unnecessary forcing. Why? Because biasing means
changing the probability of sampling in different parts of phase space, using
"weights" to correct for the unphysical probability distributions. For a number
of samplings tending to infinity, that would not matter: if the biasing is
correct, mathematically the product "weight x probability" remains constant.
But the number of samplings is not infinite: and it can happen that some
regions of phase space remain unsampled or poorly sampled. If "probability"
happens to be close to zero, whatever weight is applied the result will stay
zero. There is no strict rule about avoiding that. One needs experience, and
being prudent. In Hayley's case, increasing the interaction probability by 1.E6
is too aggressive: there is the risk to miss some other important effect.
But is 5E-4 reported by George OK? As I said, there is no hard rule. I would
avoid even that amount of biasing, but it may depend on the problem. In such
borderline cases, it is always wise to do some tests: try 1.E-4 and 1.E-3: do
results change a lot?
But after all, why do you think that we have to bias so much? Because otherwise
most particles would pass through the target without interacting. But does it
matter? In Hayley's problem, after the target there are only vacuum and
blackhole. The waste of CPU time would be minimal: no further tracking is
required. Run a very large number of primaries, without worrying about wasting
a large fraction of them.

> The manual explains the
> difference between positive and negative values of this factor. I don't think
> it would make any difference for this problem, but to be clear on the subject:

This is irrelevant here. Positive and negative values of the factor only
change the way secondaries are treated.

Target thickness:
> Now regarding the target thickness: this has also been discussed in different
> contexts. One such discussion is here:
> http://www.fluka.org/web_archive/earchive/new-fluka-discuss/2425.html

I don't like George's suggestion to replace the target by a thicker one with
the same g/cm2. Why? because it is not necessary! Are we sure that a few
micrometers are too thin for FLUKA? The link (an answer by Giuseppe
Battistoni) reported mentions a possible problem only for two reasons:

> 1) First of all the minimum thickness in a transport code like FLUKA is
> meaningful if there are enough atoms to make valid the use of multiple
> coulomb scattering approach and energy loss
> calculations. Therefore it depends on the medium, its atomic properties
> and the density.
Changing the thickness, the number of atoms stays the same: so it does not
matter.
> 2) There is at least a second consideration: in a given geometry the
> minimum thickness that can be appreciated depends on its ratio with
> respect to the maximum size of the problem itself.
In Hayley's case, there is only one relevant thickness: the few micrometers
of the target. The lateral size does not matter here.

> The idea of playing with material density has been discussed for different
> types of problems. One example I recall is gas bremsstrahlung.

That is different! In the case of gas bremsstrahlung you must suppress multiple
scattering and delta rays. Playing with material density is meaningful only if
there is no multiple scattering and no secondaries, as explained in point 1)
of Battistoni above.

> I was pretty
> confident about the approach, but H-3 production question raised some
> questions. I am comparing 2 models with the only differences in the target
> thickness and density. In one model I have 0.0001 cm target made of regular
> carbon (density 2 g/cm^3) and in the other model I have 10 cm target made of
> carbon at 2.0E-5 g/cm^3. Both are 2E-4 g/cm^2, but produce very different
> activity of H-3. I hope someone else can see what I'm doing wrong here.

What you are doing wrong is that you neglect all effects of scattering (not
only multiple scattering but also possible geometry effects: for instance,
depending on your geometry, scattered particles escaping from the target
in a different way). All effects where angles are concerned, are different
in targets of different size.

Alberto




> - George
>
>
> --
> George Kharashvili
> Jefferson Lab Radiation Control
> 757-269-6435
>
> ----- Original Message -----
> From: "hayley smith" <hayley.smith_at_stfc.ac.uk>
> To: "George Kharashvili" <georgek_at_jlab.org>
> Cc: "FLUKA Discussion List" <fluka-discuss_at_fluka.org>
> Sent: Wednesday, February 25, 2015 3:36:25 AM
> Subject: RE: [fluka-discuss]: RESNUCL with LAM-BIAS
>
> Hi George
>
> Many thanks for your reply, I really appreciate it.
> Thank you for confirming that there is no normalisation required to correct for the use of LAM-BIAS.
>
> I have spent some time understanding the changes you suggested to my input file. I was interested in the idea of redefining the material with a lower density, increased length and subsequent smaller reduction factor in the LAM-BIAS. I hadn't thought of doing this before, neither had I seen it mentioned much in the discussion list.
>
> I have spent time running your file (foil 10 cm, LAM-BIAS = 1e-2, 1e8 primaries) to compare to the results I had with my input file (foil one micron, LAM-BIAS = 5e-7, 50000 primaries). I was happy to see that for the most part they agreed remarkably well (see attached spreadsheet or picture). The main differences I observed were that initially my model created Boron-13 and at each decay time my model created Hydrogen-3. I tried modifying WHAT(5) of RADDECAY to see if this was the difference but it didn't seem to achieve agreement. Maybe it's a convergence issue? I wasn't able to run more than 1e8 particles in any useful time frame. Or maybe this is an artefact of the 'over-biasing' you mention in my input file? From my understanding the formation of tritium is important to consider.
>
> Thank you once again for your email, it has certainly given me a better understanding of different ways of using FLUKA to solve problems, and given me some confidence that my input file was producing meaningful results.
>
> Hayley
>
>
>
> -----Original Message-----
> From: owner-fluka-discuss_at_mi.infn.it [mailto:owner-fluka-discuss_at_mi.infn.it] On Behalf Of George Kharashvili
> Sent: 19 February 2015 15:49
> To: Smith, Hayley (STFC,RAL,ISIS)
> Cc: FLUKA Discussion List
> Subject: Re: [fluka-discuss]: RESNUCL with LAM-BIAS
>
> Hello Hayley,
>
> I think your foil may be too thin and your hadronic inelastic interactions - overbiased.
> You can redefine your target material(s) (using MATERIAL card) with lower density and increase the target thickness so that you end up with the same g/cm^2. Using 1E-2 in WHAT(2) of LAM-BIAS is a reasonable value from my experience. I would also use WHAT(4) to define which particles the biasing applies to (even though the default may be protons).
>
> RADDECAY - if you would like to kill both electromagnetic cascade and transport of decay radiation, set WHAT(5) of RADDECAY card to 9999999999.
>
> There is no normalization needed to correct for the use of LAM-BIAS, FLUKA takes care of it all. However, the RESNUCLE units will be Bq/cc only if you assign the correct volume of the scoring region in WHAT(6). Otherwise it is Bq per region. I'm not sure why you are scoring RESNUCLE in all regions. Don't you wont to score it in the region called FOIL?
>
> I have attached a modified version of your input file that you may find useful.
>
> Cheers
>
> - George
>
>
> ----- Original Message -----
> From: "hayley smith" <hayley.smith_at_stfc.ac.uk>
> To: "FLUKA Discussion List" <fluka-discuss_at_fluka.org>
> Sent: Monday, February 16, 2015 11:44:29 AM
> Subject: [fluka-discuss]: RESNUCL with LAM-BIAS
>
> Hi
>
> I am trying to look at the residual nuclei produced when a 70 MeV proton beam
> repeatedly passes through a thin (on the order of microns) foil. I am trying
> to understand the different residuals, and proportions of residuals, produced
> by different foil materials (carbon, aluminium etc) and I have a few questions
> about my methods:
>
> From reading many of the previous FLUKA discuss posts I have understood that
> on these short length scales it is necessary to use LAM-BIAS to reduce the
> nuclear interaction length. At 70 MeV the .out file showed that the inelastic
> interaction length was 34.31 cm in carbon so I reduced the interaction length
> by 1e-6 using the LAM-BIAS card. Is this a reasonable scaling factor?
>
> I am now looking at the RESNUCL output that I get but am confused about
> interpreting the results, even after consulting previous discussion list posts
> and the notes from the most recent course. I have tried repeating the
> simulation with LAM-BIAS for different numbers of primary particles (1000 -
> 50000) and different scaling factors to get an understanding. In my
> simulation I assume a proton distribution of 1.25e15 protons per second. I
> have looked at the .out files and .lis files but I am still unsure - do I need
> to apply any weighting to the RESNUCL output or has this already been taken
> into account when the Bq/cc are processed?
>
> And are there any other settings that I may be missing in my very simple input
> file (attached) that would allow me to get a useful simulation of the
> residuals?
>
> Many thanks for any clarification you are able to give,
>
> Hayley
>
>
>
> Hayley Smith
> ISIS Facility
> Rutherford Appleton Laboratory
> Harwell Oxford
> OX11 0QX
>
> 01235 445524
> hayley.smith_at_stfc.ac.uk<mailto:hayley.smith_at_stfc.ac.uk>

__________________________________________________________________________
You can manage unsubscription from this mailing list at https://www.fluka.org/fluka.php?id=acc_info
Received on Thu Feb 26 2015 - 17:10:14 CET

This archive was generated by hypermail 2.3.0 : Thu Feb 26 2015 - 17:10:24 CET