Dear Michael Seaton,
Thank you very much for detailed cause of error. I could not imagine this.
I see about the requirement of code development for parallelepiped system.
And also appreciate for the suggestion of alternative way. Is it physically meaningful to modify coordinates of atoms in the solid state? I am not sure whether it can be similar to real situation.
In other way, is it possible to enlarge the electronic grid far beyond the atomic cell in DL_POLY? In Darkins et al.(https://doi.org/10.1016/j.commatsci.2018.02.006), they wrote "it is common to extend the electronic grid far beyond the MD cell in the perpendicular (x and y) directions. This extended electron grid enables the electronic energy to be transported far from the volume of the initial deposition, thus simulating electronic heat transport into the bulk material". I would like to ask it can be the alternative way to solve it.
Thanks in advance, Kim.
-----Original Message-----
From: Michael Seaton <[log in to unmask]>
Sent: Wednesday, April 10, 2019 12:19 AM
To: [log in to unmask]; Kim <[log in to unmask]>
Cc: Michael Seaton <[log in to unmask]>
Subject: Re: Error when TTM applied to large system
Dear Kim,
Apologies for taking so long over this, but I have finally found out why your system fails when using TTM.
The implementation of TTM I have included in DL_POLY_4 assumes that the atomistic system is orthorhombic (cubic or cuboidal). However, the CONFIG file you have supplied is for a monoclinic parallelepiped system with some particles' x-coordinates extending beyond the values one would expect for a cuboid of the same dimensions.
While this is not a problem for the main atomistic simulation, it is a problem for TTM as its current implementation does not allow the temperature cells to be shaped as parallelepipeds. Because of this, some particles in each processor core lie outside the defined temperature grids. This leads to array out-of-bounds errors, particularly when calculating Langevin forces (which is the first time these grids are used in a Velocity Verlet calculation).
A lot of further code development work would be required to get DL_POLY's TTM implementation to work with parallelepiped systems. (I would like to do so, but this is likely to take some time due to the large number of modifications required to determine the voxel for each particle based on their positions and to carry out the thermal diffusion calculations.)
However, if you wish to get your system to work immediately, I would recommend modifying the configuration to give a cuboidal box: since the system is monoclinic, this should be reasonably straightforward as only the x-coordinates of some atoms need to be modified. If you cannot or do not want to reopen the CONFIG file in OVITO, I can come up with a Python script to modify those coordinates for you if you are interested?
Regards,
Michael
########################################################################
To unsubscribe from the DLPOLY list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=DLPOLY&A=1
|