The ABINIT Users Mailing List ( CLOSED )

[Steven Homolya <Steven.Homolya@spme.monash.edu.au>]

I am a fairly inexperienced user of abinit, but there are several problems
I can see with your input file:

1. Most important one:

toldfe 1.0d-6

will stop scf calcn. when total energy has converged to 1e-6 Ha;  usually
not good enough to get precise forces. You'll need to converge for force
instead; see input variables toldff and tolmxf. Former should be 10 times 
smaller than the latter. This is covered in the tutorials.

2. ionmov 3 is, in general, better than ionmov 2, because it will monitor
your total energy, trying to make sure it is decreasing. This prevents 
numerical divergence for bad initial guesses and the parameter space is 
searched more "intelligently". See description of input variable ionmov.

3. Not related to your qn. but important for efficiency / accuracy:

ngkpt along a reciprocal primitive lattice vector should be proportional
to the length of the prim. latt. vector (as far as possible with integer
ngkpt). This will ensure a more "even" distribution of k-points in 3-D.  
For your case (with orthogonal rprim's) it's particularly simple.  You
need

ngkpt X Y X

where

X = round( Y * 14.683171 / 10.365147 )

because

> acell  10.365147  14.683171 10.436957

E.g.,

ngkpt 6 4 6
ngkpt 4 3 4

... if I understand the definition of ngkpt correctly.

Steve

--
Steven Homolya
School of Physics and Materials Engineering
Monash University, VIC 3800
Australia
Tel: +61 3 9905 3694
Fax: +61 3 9905 3637