[ESPResSo] local pressure tensors

[Benedict Reynolds]

Hi all,

I'm planning on committing changes that involve the calculation of local
pressure tensors.

Currently we have two routines that calculate stess tensors: "analyze
p_IK1" and "analyze stress_tensor".  "analyze stress_tensor" returns the
global stress_tensor. "analyze p_IK1" is given a list of particles, a
box volume, and a flag.  It should be able to calcualte a global or a
local stress tensor.

As far as I can tell "analyze stress_tensor" works fine. "analyze p_IK1"
works fine when the box volume is set to the simulation box volume, the
list of particles corresponds to a list of all particles in the
simulation, and the flag is set to 0.  It does not, however, give
sensible results if one is trying to find a local pressure tensor.  The
problems are as follows:

- bonded interactions between a particle in the list and a particle out
of the list are only included when the flag is set to 1 and if the bond
information is stored on the particle in the list.

- when the flag is set to 0, non-bonded interactions between particles
in the list and not in the list are not included.  When the flag is set
to 1, non-bonded interactions between two particles, both of whom are in
the list are not-included.

- all interactions are either included in the pressure tensor entirely
or not at all.

 - also if one if trying to find the pressure tensor in many bins, it is
much more efficient to be able to do these operations simultaneously.

My suggestion is that we have a command "analyze local_stess_tensors".
Which calculates local stress tensors for cuboid bins in a grid, where
the size and position of the grid and the number of bins in each
directions is specified.  Interactions contribute to the stress tensor
in a bin according to what fraction of the line connecting the two
particles passes through the line.

I'd also be interested to know if any people use p_IK1.  My preference
would be to entirely remove it since I don't think it has any
functioning functionality that "analyze stress_tensor" doesn't have.  Or
perhaps I could leave it in for backwards compatibility but display a
warning message if it's being used in a way that could produce
misleading results.

Any thoughts?

Cheers,
Ben