h5md-user
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: H5MD proposal for quaternion


From: Pierre de Buyl
Subject: Re: H5MD proposal for quaternion
Date: Fri, 8 May 2020 22:52:01 +0200
User-agent: NeoMutt/20180716

Hi all,

On Tue, Feb 11, 2020 at 05:28:05PM +0100, Konrad Hinsen wrote:
> > Some elements already depend on the creator. This is the case for charge, 
> > even
> > though H5MD considers some typical scenarios already, for species, and for 
> > ID.
> 
> True, but charge still has more of a universal meaning than quaternions,
> which can represent orientations or rotations, and in the former case
> require a reference conformation that should ideally be deducible from
> the information in the trajectory.
> 
> > I would be in favor, at least for quaternion, to restrict the definition. 
> > For
> > instance: "The quaternion defines the orientation of the particle as 
> > follows.
> > Any point on the body represented by the particle, and located at "p0" at an
> > arbitrary reference in time, is located at the time corresponding to 
> > quaternion
> > q1 at p1 = q1 p0 q1* where q1* is the conjugated quaternion to q1 and the
> > multiplications in the formula are quaternion products. The position of the
> > point so defined is always relative to the position of the particle."
> 
> OK, that's a lot more precise, but wouldn't this better be called
> "orientation"?

The "physical" meaning has more consistency overall, I'll switch it. It is funny
though, as in "MD simulation world", I believe that quaternion is rather
unambiguous :-)

> > Whatever happens with quaternions, you propose to have a generic ways of 
> > storing
> > creator-specific elements, is that right? Do you have an idea of how that 
> > would
> > work out?
> 
> Yes, that's what I have in mind. The idea is to use H5MD much like XML
> or HDF5, as a foundation of basic data structures to which more specific
> data definitions can be added. For example, H5MD defines how to store
> scalars, vectors, and quaternions, all of that time-dependent or not,
> plus of course the most basic geometric data that the definition already
> contains. A creator-specific definition of "orientation", for example,
> would then just add "H5MD time-dependent quaternion per particle" plus
> the semantics for interpreting the quaternions.

This is a nice idea for a H5MD 2.0. If there was a way to fund this, I would
follow on it. In the light of career perspectives, this will likely not happen
(not metaphysical here, practical).

In the interest of feasibility, I'll carry on with the current proposal to add a
specific orientation item.

Proposed updates:

1. A "length-4" orientation means quaternion, such that the orientation vector
is u = q u0 q* with the usual quaternion product used on the right-hand side.

1. A "length-D" (where D is dimension of position dataset) means -> the unit
vector giving the orientation.

I updated the proposal, it should appear online soon.

Cheers,

Pierre





reply via email to

[Prev in Thread] Current Thread [Next in Thread]