Hi,

> F = exp(A), and A =
>
>  a  b  c  d
>  b  a  u  v
>  c -u  a  w
>  d -v -w a
>
> This seems to be the case for all matrices that we are concerned about in
> the propagation parts of ARTS, even in the scattering cases.  Is this true?
>

After a bit of reading (e.g. del Toro Iniesta's "Spectropolarimetry"), I
come to the conclusion that all RT propagation matrices (i.e. the K)
exhibit this structure. It's an effect of the geometry relations between
electric/magnetic field and polarization frame. And it's valid for
individual (wave) K as well as the "bulk" K (over different medium
constituents and wave interaction processes).


Got me thinking, though, that we are wasting a lot of memory and a lot of
> computing time keeping the entire 4X4 propagation-matrix [...] rather use
> Vectors of these parameters to represent the matrix.  However, that would
> be overly simplistic and might not be beneficiary enough to justify the
> extra work.
>

Seems a clean solution to redefine ext_mat (and similar containers) to only
hold the (max) 7-element vector.

As Patrick mentioned, for scattering data we do something similar already.
at least for the input data (when needed in the RT solution, the data are
currently also converted to the up to 4x4 matrix).

The general particle class is not implemented, i.e. the 7-element format is
not fixed. for the azimuthally random case we use order Kjj, K12, K34 (or,
for the A definition above: a, b, (-)w). This, however, is not directly
applicable for the all-sky RT ext_mat format since that is rather governed
by the stokes dimensionality instead of symmetry relations (e.g., for
stokes_dim=2 we don't need the K34 while for stokes_dim>2 we need
additional elements that are 0 in the azimuthally random scattering
particles case).

>From the view point of elements we need for the different stokes_dim
levels, format should be like [ a, | b, | c,u, | d,v,w ], i think (the
vertical dashes indicate the limits for the different stokes_dim, i.e. the
reduced ext_mat would have length 1,2,4, or 7 for stokes_dim=1,2,3,4
respectively).



>
> Instead, I would like to propose a similar class, PropagationMatrix, that
> can store the entire propagation matrix in parameterized form that is
> reduced to the seven variables above by simple mechanisms.
>

I can not judge, whether it's advantageous to further parametrize the K
contributions from different physical processes. In any case, I don't see
that this would be mutually exclusive with a matrix-to-vector format change
of ext_mat (since, as written above, ANY K has to exhibit the structure of
A, i.e. all parametrizations anyway happen only on the 7 elements
individually, not on the 4x4 elements (or, differently expressed K34 will
always be equal -K43, independent of any parametrization of the different
processes and in different conditions, i.e. can always be expressed by just
one value together).

So, if that is advantageous, there could rather be different propmat
classes, depending on the process. that in the RT solution (by the
abs_xsec_agenda or so, i guess) than would need to be combined into one
common ext_mat vector.


Regarding the scattering matrix issues, Robin & Patrick brought up, I think
we should discuss this independent of the propagation matrix issues here.
hence, i'm taking this in a separate mail...

wishes,
Jana


-- 
=====================================================================
Jana Mendrok, Ph.D. (Researcher)
Chalmers University of Technology
Department of Space, Earth and Environment
SE-412 96 Gothenburg, Sweden

Email: [email protected]
Phone : +46 (0)31 772 1883
=====================================================================
_______________________________________________
arts_dev.mi mailing list
[email protected]
https://mailman.rrz.uni-hamburg.de/mailman/listinfo/arts_dev.mi

Reply via email to