[h5md-user] thermodynamics module

[Felix Höfling]

Hi all,

Peter, thanks for creating the thermodynamics module! A few amendments and  
remarks from my side:

I added the volume which was missing as the variable conjugate to the  
pressure.

I think we should go for either intensive or extensive quantities, this  
makes life easier for analysis tools. It is not a big deal for the writer  
to convert the data (simply multiply or divide by the particle number).  
Further, people may also work with, e.g., the energy density (per volume).  
And one may also want to store the (fluctuating) density, and so on.  
Actually, I have already made a choice for extensive quantities (when  
applicable, e.g., pressure and temperature are intensive by their nature)  
and dropped the per-particle averages.

I wondered whether 'dimension' and 'particles' should become mandatory to  
ease the conversion.

Further, 'dimension' is not really an H5MD 'element' as it does not change  
in time, it carries no unit etc. It could become an attribute as it is for  
the box.

For anisotropic barostats, one may want to store the stress tensor instead  
of the pressure. I'm not sure what the conjugate variable would be then,  
probably the matrix of edge vectors of the volume. We can leave this for  
future extensions.

The possibly most controversial point: I'm not happy with the wording

    For each subsystem stored in the particles group as /particles/<group>,
    the observables group shall contain a corresponding group
    /observables/<group> with one or more of the following elements:

It means that for each group in particles there _must_ be a group in  
observables and vice versa (even if the subgroup in particles is empty,  
but it must exist). I would like to avoid such an interdependency between  
different root groups.

Second, there is an inconsistency between the core definition of  
"observables" and the thermodynamics module: the former allows H5MD  
elements in "observables" itself, the latter not. I would like to rephrase  
very generically as follows:

    The `observables` group or any of its subgroups may contain one or  
more of the following elements:

(We don't need to repeat that the subgroups refer to subsystems, this is  
in the core definition.)

I'm happy to receive for your feedback.

Regards,

Felix