Previous Thread
Next Thread
Print Thread
Periodicity in 1 or 2 dimensions
#975 02/23/04 11:07 PM
Joined: Oct 2003
Posts: 6
jfm17 Offline OP
Forum Member
OP Offline
Forum Member
Joined: Oct 2003
Posts: 6
2 questions related to periodicity in 1 or 2 dimensions:
I tried a lattice minimization (based on the test case) on a crystal structure of a polymer, using image patching across the boundary of the unit cell. The ABNR minimization exited before reaching the maximum nstep, and the first derivative was unchanged for any of the atoms (far from zero). Does lattice minimization not work with image patching, or am I missing something?

Also, I've come across some papers that derive Ewald summation methods for systems that are periodic in 2 dimensions (such as a solid surface or membrane) or periodic in 1 dimension (a pore or channel). Both of these papers state that the "usual" 3D Ewald is not appropriate for this type of periodicity, but do not explain why it does not work. Applying 3D crystal symmetry leads to lamellar stacking of surfaces or parallel helices or pores - and I believe this is the only way CHARMM can handle these types of periodicity using Ewald electrostatics. Can someone with a better understanding explain why this is inappropriate when using 3D Ewald, or alternatively why these authors are overstating their claims?
M. Kawata et al J. Chem. Phys. 116 (8) 3430-3448 February 2002
A. Brodka Chem. Phys. Letters 363 604-609 September 2002

Thanks,
james

Re: Periodicity in 1 or 2 dimensions
jfm17 #976 03/09/04 10:20 PM
Joined: Feb 2004
Posts: 13
Forum Member
Offline
Forum Member
Joined: Feb 2004
Posts: 13
2D or 1D ewald summation is not built into CHARMM. Although I imagine you could change this with modifications to the source code if you have the academic version.

I needed to use 2D periodicity in a recent surface simulation. In my case, I had a polyelectrolyte adsorbed on a surface. Using 3D ewald summation would have been inappropriate, since my macromolecule would be interacting with images of itself in the vertical direction, and that would be an artifact not representative of my actual system.

In order to get around this, I used a psuedo 2D boundary condition. I did this by placing 30 Angstroms of water above my macromolecule, placing a "lid" above that (a surface of fixed atoms) and placing 200 angstroms of vacuum above that. (The lid kept the solvent from evaporating into the vacuum). This may have been overkill, but I wanted to make sure to minimize any contribution to the ewald summation in the z direction.

Regards,

Jim Sullivan


Molecular Engineering Laboratory University of Washington
Re: Periodicity in 1 or 2 dimensions
jfm17 #977 03/09/04 11:23 PM
Joined: Sep 2003
Posts: 394
Forum Member
Offline
Forum Member
Joined: Sep 2003
Posts: 394
The 0 dimension periodicity (i.e. NlogN no cutoff method) has been an undocumented option for several years. I should probably document it. It's the PME FINIte option (see the source code in nbonds/nbutil.src for info).

We have been implementing the 2 dimension periodicity (but no working code yet). This should be available in the near future.

Regarding the lattice minimization: I think you may be misreading the documentation. Do you want to minimize only the unit cell dimensions? or do you also want to minimize the coordinates at the same time?

Re: Periodicity in 1 or 2 dimensions
BRBrooks #978 06/11/04 06:34 PM
Joined: Oct 2003
Posts: 6
jfm17 Offline OP
Forum Member
OP Offline
Forum Member
Joined: Oct 2003
Posts: 6
It's exciting to hear the 2D PME is under development... The lattice minimization is still not working. I've tried using the "nocoordinate" option and it didn't work either. This seems to be a problem with the image patching. My polymer is cellulose - a flat chain of glucose rings that is 8 residues long, with the first and last residues bonded to each other across the periodic boundary. The image psf shows the proper atoms bonded to each other, but when I run a regular minimization (not lattice) the end residues flip so they are nearly parallel to the edge of the box. The interior 6 residues are stretched as if they were the ones bonded across the boundary, but they are not according to the image psf. I am unsure how the count of exclusions is calculated -- I am using nbxmod 5 (or -5) and it looks like exclusions are appearing for atoms separated by more than 3 bonds. I'll put together a file that will reproduce this and post a link.

james

Re: Periodicity in 1 or 2 dimensions
jfm17 #979 06/11/04 09:28 PM
Joined: Oct 2003
Posts: 6
jfm17 Offline OP
Forum Member
OP Offline
Forum Member
Joined: Oct 2003
Posts: 6
here is a tar file that contains everything to reproduce this problem... any comments or suggestions would be welcomed!
Thanks,
james

http://www.people.cornell.edu/pages/jfm17/image_patch.tgz


Moderated by  lennart, rmv 

Link Copied to Clipboard
Powered by UBB.threads™ PHP Forum Software 7.7.4
(Release build 20200307)
Responsive Width:

PHP: 5.6.33-0+deb8u1 Page Time: 0.012s Queries: 24 (0.005s) Memory: 0.9244 MB (Peak: 1.0126 MB) Data Comp: Off Server Time: 2020-10-29 07:56:44 UTC
Valid HTML 5 and Valid CSS