Removed triclinic restrictions on slab ewald/pppm.

[frobnitzem]

Since the slab correction applies specifically to the z-direction, it is not influenced by tilt. I have confirmed the correction by calculating it manually for a system with 2 charges (+1 at 0,0,1.5 and -1 at 0.0,-1.5). The files are attached:

ewtest.zip

Four cases are tested -- an orthogonal and a triclinic box with and without the slab correction.
Although changing the box tilt alters the energies and forces, the correction term is identical for both, as it should be.

[stanmoore1]

Unfortunately, it isn't this simple. This PR fails the most basic sanity check. An orthogonal box should give the same energy and force as a triclinic box when all the tilt factors are equal to 0.0, since the triclinic box isn't really tilted. However, if I add this line below to an input script using this PR I get significantly different forces and energies than the orthogonal case:

change_box all triclinic xy final 0.0 xz final 0.0 yz final 0.0

A second good check for triclinic systems is that when there are periodic boundary conditions in x and y, the system energy and force for an orthogonal box will match a triclinic box with tilt factors xy = Ly and xz = Lz, where Ly is the total box length in the y direction and Lz is the total box length in the z direction, i.e.:

change_box all triclinic xy final ${Ly} xz final ${Lz} yz final 0.0

A third good check for kspace slab systems is that this input script and data file (attached below) give the analytic solution of forces in magnitude equal to pi (3.14159...), see the book Understanding Molecular Simulation, Frenkel and Smit, p. 318:

data.two.txt
in.test.txt

[stanmoore1]

For the Kspace slab analytic solution, just place two oppositely charged particles in a slab system. They form two infinite sheets of charge due to periodic boundary conditions in the x and y directions. If the charges are far apart enough in the z-direction, the force in the z-direction for the orthogonal case converges to 2*pi*q^2/(Lx*Ly)

[stanmoore1]

Sorry, in the above post that should be tilt factor = 0.0, not 1.0. I've edited the post above. If I run the script I attached with a pure orthogonal box, I get an energy = 6.5933251. If I run it with triclinic and all zero tilt factors, I get an energy = 2.6182985. The energies should match in these two cases but don't.

[frobnitzem]

This is certainly odd behavior. I can confirm it even with your first example. The force is radically different even by setting the box using: run 0 print "f = ${fz}" change_box all triclinic xy final 0.0001 xz final 0.0 yz final 0.0 run 0 print "f = ${fz}" It's definitely related to the extension in the z-direction, since when I set "kspace_modify slab 1.0". the forces before and after switching to triclinic are nearly identical: f = 3.14494261727358 f = 3.1449426178097

…

On 3/11/17 9:19 PM, Stan Moore wrote: Sorry, in the above post that should be tilt factor = 0.0, not 1.0. I've edited the post above. If I run the script I attached with a pure orthogonal box, I get an energy = 6.5933251. If I run it with triclinic and all zero tilt factors, I get an energy = 2.6182985. The energies should match in these two cases but don't. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#411 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACI7d7yrbNikO1oeu6WLQc9UMS4D1i7bks5rk1Y8gaJpZM4MaEec>.

[akohlmey]

i looked into this a little bit today. turns out, the slab correction does work correctly with the provided modification, but that is not the origin of the difference. the issue is, that activating the slab correction changes the volume of the cell and that means, that all computations done before applying the slab correction, need to account for that correctly, and that is where things get a bit complicated, since for triclinic cells, all of those computations are done in lamda coordinates.
thus i will close this pull request as "invalid" for now, and we only keep issue #410 as a reminder to look into it, when somebody has the time.