RE: [fluka-discuss]: low density question

From: Alexander MacKinnon <>
Date: Wed, 27 Nov 2013 16:45:36 +0000

Dear Francesco,

thanks for the very helpful reply. We'll pay more attention to the Sternheimer parameters, hopefully to overriding them in a way that's appropriate to an ionised medium. And thank you for pointing out that most of our regions are being treated as vacuum! As you say there are clues in the .out file but there is so little column density in each of those layers that we might not have realised this was happening. Smaller, denser slices of material will tell us what we want to know.

Best wishes,

Dr Alexander L MacKinnon
Centre for Open Studies
University of Glasgow
11 Eldon Street
(+44)(0)141 330 1857
From: Francesco Cerutti []
Sent: 24 November 2013 20:14
To: Alexander MacKinnon
Subject: Re: [fluka-discuss]: low density question

Dear Alec,

your diagnosis is pretty correct, the crash is due to the terribly low
density value, as used in the Sternheimer parameterization of the density
effect correction to stopping power. But please note that values below
10^(-10) g/cm^3, far from working well, are actually meaningless, since
the code in this case just resets the respective material to vacuum
(issuing a "material changed" message - by the way incomplete - in the
output file).

Kind regards


Francesco Cerutti
CH-1211 Geneva 23
tel. ++41 22 7678962
fax ++41 22 7668854

On Fri, 22 Nov 2013, Alexander MacKinnon wrote:

> Dear colleagues,
> I've been experimenting with FLUKA as a tool for particle transport in an astrophysical setting. I've followed to some extent the example of the cards for studying cosmic rays in the terrestrial atmosphere but have run up against an interesting problem. Our lengthscales are very large and our densities very small, by most people's standards, and I think it's from density that the problem arises.
> The attached .inp file attempts to implement a target with successive layers of monotonically increasing density, according to one of the models for solar atmospheric structure. With this .inp file FLUKA fails without producing any .out, .log or .err files. The temporary versions of the .out and .log files do exist, however, in the fluka_XXXX folder that doesn't get deleted when there's a crash. The .log file from there tells us there's a floating point exception in a function called dpdx.f which I guess calculates the energy loss rate.
> By experiment we discovered that this only happens if one of the layers has a density in a well-defined range, between 10^(-10) and 1.82 10^(-10)^(-3) . As long as all densities are greater or less than this value FLUKA runs without crashing.
> There are some ways round this problem. Only the total column density in each of the layers really matters for most purposes, so we could easily generate a different density structure that would tell us what we need to know, avoiding the problematic density range, and then re-apply the results in the real atmospheric structure. For some purposes we only need thick target yields or secondaries emergent at the top of the atmosphere and again the detailed density structure would not be important. Etc. But we'd prefer to understand what's happening in case this highlights a bigger issue we should be aware of.
> Thanks in advance for advice.
> Regards,
> Alec MacKinnon
> PS let me anticipate some more general comments. The higher layers of our atmosphere are of course not neutral but there's very little column depth there. The solar atmosphere is almost neutral throughout a large fraction of the stopping depth of the particles of interest. Results from a code like FLUKA give a valuable reference case even if there are significant plasma physics complications in transport.
> Dr Alexander L MacKinnon
> Centre for Open Studies
> University of Glasgow
> 11 Eldon Street
> UK
> (+44)(0)141 330 1857
Received on Wed Nov 27 2013 - 18:45:02 CET

This archive was generated by hypermail 2.3.0 : Wed Nov 27 2013 - 18:45:02 CET