Emma,
first of all, to be sure about your coordinates, just draw crystal using
your .xv file generated by SIESTA. Use, for example, Andrei's tool
siesta2xcrysden (see in folder /Util/Contrib/APostnikov that comes with
SIESTA 2) and www.xcrysden.org to visualize it. f by repeating your unit
cell by given translation vectors you will see the needed cell, then
its's OK.
A couple of things to point out:
(1) the forces on the atoms in the unit cell are HIGH (0.9 eV/Ang) but
only in Z.
DFT cannot give you the experimental lattice constant.
This is not a cubic lattice, thus you need to treat independently c and
a lattice constants. Setting just lattice constant 3.96 Ang as you did
is not enough, you also need to vary the c/a ratio (which you fixed at
1.679).
The "simplest" way to do this, is to perform VariableLattice
optimization of the cell - it will relax bith c and a, although you will
need to play with it several times, since usually it will also move the
atoms in the cell from symmetric positions.
(2) when I vary the lattice parameter slightly (as if I was calculating
the cohesive curve to get the bulk modulus) I do NOT see the expected
parabolic dependence of the energy on the lattice constant (with a
minimum at the equilibrium lattice constant). Instead I see an increase
in the energy with increased lattice constant. (although I do see an
approximately linear decrease in the pressure with increasing lattice
constant).
Usually this is an indication of small energy cutoff.
100 Ry is definitely not enough, since you use NLCC (needed cutoff will
also depend on your pseudopotentials cutoffs). Ideally you need to check
the convergence of your energy vs. energy cutoff, k-grid, basis.
CoI2 is an ionic material?, so if you want to get more accurate results,
I would recommend to use GGA instead of LDA. And this would require even
higher energy cutoff.
Sincerely,
Alexander
