Date: Thu, 12 Jul 2001 12:15:46 -0700 (PDT)
From: Thomas Cheatham
Subject: Re: ptraj diffusion


> I have some questions about the output of the diffusion option
> in ptraj.

The diffusion command in ptraj basically calculates the mean-squared
distance vs time, or equivalently, for each frame, the squared difference
of the coordinates to their starting position (*). The distance, in
general, is in angstroms**2 and the time depends on your trajectory frame
sampling rate (the default is 1.0 ps between frames). You can set the
time (in ps) with the time keyword, i.e.

diffusion :WAT average time 0.5

...will calculate the average mean-squared displacement vs. time for all
residues named WAT assuming a time interval of 0.5 ps between frames in the
trajectory...

If you take this data and plot it, the slope is A**2/ps which you can
convert to a diffusion constant by converting to cm**2/s (* 10.0) and
dividing by 6 (for the degrees of freedom).

With the "average" keyword, this mean-squared displacement is averaged
over all selected atoms at each frame of the trajectory. Without this,
the information dumped is for the average at each frame is along with
information accumulated for each atom at each frame.

(*) An additional complication is that in many simulations, imaging of the
trajectory is performed on the fly. Therefore, in order to account for
this, the diffusion routine (as described in the
transformDiffusion() subroutine in actions.c) "unimages". It does this by
checking to see if an atom moved more than 1/2 the box length between
frames. If it did, then it subtracts off the box length.

Some CAVEATS of this procedure:

(1) if an atom moves > 1 box length between sequential frames in the
trajectory, the "unimaging" will only account for movement of 1 unit
cell, not multiples. I will fix this.

(2) I just realized that I have NOT updated this code to handle
non-orthorhombic unit cells. This means that the UNimaging procedure
is broken for truncated octahedron (if imaging was performed during
dynamics, i.e. IWRAP=1) or any unit cell that is not 90x90x90
degrees... I will also fix this.

> First, I presume that the columns signify different atoms,
> but for some reason the number of columns is different than
> the number of atoms in the mask (408 instead of 553, e.g.)
> when I open the .xmgr file in SigmaPlot.

Offline, we can try to track this down if you send me more
information. First off, make sure that ptraj recognized your atom
selection. Note that in general with ptraj, at least with newer versions,
you can set a variable called prnlev to a higher value and dump all sorts
of debugging information, i.e. in the ptraj input:

prnlev 4

The most recent version of ptraj is accessible at

http://www.chpc.utah.edu/~cheatham/software.html


If you or others have further questions regarding ptraj (bugs, wish
list for features, how to use it) please feel free to e-mail me directly
or ask the list,

--tom