supercells and how abinit does them

[joe]

Hello,
something that has been driving me crazy... I cannot find anyone that discusses explicitly how the supercell is created in abinit (my apologies if I missed it somewhere). I am working on nanotubes, but my question is very general.

Question: How does the supercell relate to the acell? I have read the documentation and it still is illuding me. Another way to phrase this question is to ask, when I include more k-points in my calculation, is that like saying that I have images of my unit cell further out? Usually I think of k points as coming from the reciprocal of the real lattice, so if I have more kpts than atoms, am I using a large supercell?

I want to make my nanotube infinite in length, but it is unclear how this is done in the context of a supercell.
any input would be much appreciated.

[fariba]

Hi Joe
first : If you have large number of atoms , you shouldn't use large kpoints.
it makes your job so slow or stops it.
2nd : The code will repeat your supercell in 3 direction,so set 2 lengths
of supercell(not lie in central axis of nanotube) much larger than it is,
and let the other real.

be lucky
fariba

[joe]

Thanks much for the reply. Perhaps this is asking a lot, but I need more details I think. If I have a slice of the tube along the z axis (I have a small set of coordiantes), yes I put extra space in the x and y, but how do I ensure that the slice of the tube is replicated in the z direction to replicate the atoms I have in my slice (ie how do I mimick the existence of atoms which I do not give explicity, or how does Abinit do it?).
Thanks in advanced

[JEJohns]

I want to make my nanotube infinite in length, but it is unclear how this is done in the context of a supercell.
any input would be much appreciated.

Joe,
I think I can help clarify. Abinit does planewave based DFT, using periodic boundary conditions. This means that even if you do a calculation at just the gamma point, you are in effect doing a unit cell which is repeated in X Y & Z infinitely. This is true regardless of the number of k-points you sample, your calculation just won't be any good if you're dealing with a 3D system. The way to setup a super cell is explained pretty nicely in the tutorial on the main site for calculating the surface energy of Al(001). I believe it's tutorial 4. The way to setup a supercell for your tube is to leave vacuum spacing layers in the x & y directions. For example, if you had a tube that was 5 angstroms in diameter, and a unit cell length of 2.46 angstrom, you might setup a unit cell like
acell 15 15 2.46

when you enter the positions of your atoms, they might be something like
xangst 0 0 0
0 2.5 1.23
2.5 0 1.23 ....
or whatever they happen to be. Since you only cared about electronic coupling in the z direction, you might setup a kpoint grid like
nshiftk 1
shiftk 0 0 0.5
ngkpt 1 1 20
where the Kz direction of ngkpt you'd have to do convergence studies on. Similarly, you'd have to do convergence studies on the unit cell spacing to make sure you weren't getting spurious electronic coupling of one tube to its Born von Karmon neighbor. And don't worry about having more kpts than atoms, as that doesn't "mean" anything. More kpts just mean a finer sampling of the brouillon zone. Hope this helps
--James