The ABINIT Users Mailing List ( CLOSED )

[dick@fhi-berlin.mpg.de]

Dear all,

I have encountered inconsistency between equilibrium lattice parameters 
obtained using built-in lattice-optimization procedure (optcell=1, ionmov=2) 
and direct approach based on Murnaghan equation fit.

Specifically, i perform lattice optimization of eta-phase MnN, that has 
NaCl-like lattice structure. I restrict the description of the problem to 
ideal fcc-unit cell and nonmagnetic system. The same problem, however, exists 
for different magnetic configurations (FM, aFM), more realistic crystal 
structure (that is very similar to ideal NaCl), and PBE XC-functional.

First, i have used lattice-optimization routine with optcell=1 (change acell 
only) with ionmov=2. The input file reads as follows:

************begin**************
#Definition of the unit cell
acell 3*7.65
rprim  0.0  0.5  0.5
      0.5  0.0  0.5
      0.5  0.5  0.0
chkprim 0

#Definition of occupation numbers
occopt 3
tsmear 0.003675  # H
fband 3

#Definition of the atom types
ntypat 2
znucl 25 7

#Definition of the atoms
natom 2                typat 1 2     xred         0.00  0.00  0.00
   0.50  0.50  0.50

#Definition of the planewave basis set
ecut  25.0

#Exchange-correlation functional
ixc 7

#Definition of the k-point grid
kptopt 1
ngkpt 7 7 7
nshiftk 1
shiftk 0.5 0.5 0.5

#Definition of the SCF procedure
nstep 100
toldfe 1.0d-8

#Definition of the structure optimization
optcell 1
ecutsm 0.5
ionmov 2
ntime 20
tolmxf 1.0d-5
dilatmx 1.4
**************end************************

The cutoff energy and k-point mesh here are converged. I have used elevated 
"dilatmx", also smaller values provide the same results. Increase of the SCF 
run accuracy or change of the initial lattice parameter also leave results 
unchanged.

With this input file the converged lattice constant is 7.525Bohr.

To verify this result i have performed conventional lattice constant 
optimization. I have run through series of calculations with fixed lattice 
constant. The range was from 7.35 to 8.05Bohr, and the input files were 
identical to one described above (with optcell=0 :)). Obtained "total_energy 
vs. lattice_constant" points i have fitted to Murhaghan equation. The curve 
perfectly fits these points, and obtained equilibrium lattice constant is 
7.641Bohr.

Thus, only changing lattice-optimization method i have obtained 1.5% 
difference  in the equlibrium lattice constant. Furthermore, the total 
energies at the minimum are quite different:  -32.221639H vs -32.224727H, for 
built-in and Murnaghan fit respectively. All together it looks that optcell=1 
has found some artificial local minimum.

I believe that i have missed something important about built-in structure 
optimization, but i failed to figure out the problem. I would appreciate any 
hints and suggestions from your site. Thanks in advance,

With best,
Alexey Dick

P.S.:
ABINIT version info: sequential 4.2.4 PGI-compiled version taken from abinit 
web-site.
PSP info: self-made Troullier-Martins pseudopotentials for both Mn and N. 
Quality has been accurately tested.