[yt-users] Computing fields derived from projected quantities

Matthew Turk matthewturk at gmail.com
Fri Nov 19 07:07:40 PST 2010


Hi Chris,

I believe this is possible.

I think that we can possibly break this down into a few issues:

1) Obtaining "projected" quantities.  This term is used in the paper,
but it's not clear to me how one would generate those values from a
simulation.  I see a couple options.

Typically when we talk about projections in yt, we're actually talking
about line integrals that are sampled exactly once at every cell along
every possible line of sight.  To make an image, these are then
pixelized.  (In the method paper this is covered in some detail.)  The
usual issue with this is that taking, say, the line integral of
Temperature:

\int_0^L T dL'

doesn't really mean much.  The same, I would say, is true for
velocity.  Taking the line integral of velocity doesn't really have a
clear meaning to me.  So what we usually do is take some kind of
average value:

\int_0^L T w dL'

and then divide by the weighting field's integral:

\int_0^L w dL'

Usually we use density for this, because it provides a
resolution-independent sampling, which can then be deposited into a
fixed grid for an image.

So, the question of how to calculate the projected quantities looked
at in that paper is of interest to me.  What do you think?

2) The luminosity weighting of the averages over the FOV is not
problematic.  You could generate a fixed resolution buffer from a
projection, or you could create a covering grid and sum along an axis.
 Additionally this can be done off-axis with the volume renderer
without too much trouble.

3) Operating on the 2D buffer should be straighforward, but it is not
currently implemented.  I had put in radial profile analysis of 2D
images once upon a time, but they did not ever see any use and I
removed them not that long ago.  It was very straightforward to create
radial profiles on 2D images, so that won't be tricky.  Even barring
that, calculating a global average over a 2D object is pretty simple.

Anyway, let me know what you think, and if you run into any problems
please hit back to the list!

-MAtt

On Mon, Nov 15, 2010 at 1:51 PM, Christopher Moody <cemoody at ucsc.edu> wrote:
> Howdy folks,
>    I'd like to calculate the 'lambda' field parameter: <r|v|> / ( r
> sqrt(v^2+sigma^2) )  (Emsellem et. al.
> http://arxiv.org/abs/astro-ph/0703531). All of the quantities in question
> are viewed in projection - as an observer would see. As a result, my lambda
> field depends on the 'camera location' and requires the projected velocity
> and dispersion fields, as well as the projected radius, to be computed.
>
>   Is this possible, or rather, how much elbow grease do I have to put in to
> calculate this stuff?
>
> Thanks,
> chris
>
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>



More information about the yt-users mailing list