Skip to Content.
Sympa Menu

forum - cell relaxation: how good?

forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

cell relaxation: how good?


Chronological Thread 
  • From: Linhui Ye <yelinhui1@yahoo.com>
  • To: forum@abinit.org
  • Subject: cell relaxation: how good?
  • Date: Tue, 25 Mar 2003 06:00:40 -0800 (PST)

Dear abinit users,

I'm using abinit to calculate the lattice parameters
of hexagonal garphite. The 3 unit vectors are to be
relaxed(optcell=2 and ionmov=2). I have the following
questions:

1. The convergence parameters for cell relaxation, are
toldmxf and strfact, right? In cell relaxation, the
modification of any one of the two parameters is
sufficient to define a convergence criterion, right?

2. If I set strfact=1d-4 and toldmxf=1d-8, are these
resonable values? Is this equivalent to strfact=1d-2
and toldmxf=1d-10?

3. The output shows(only diagonal element shown here):
sigma(1 1)= -1.66117688E-05 (Hartree/Bohr^3)
sigma(2 2)= -1.66117688E-05
sigma(3 3)= 5.92238033E-06
- sigma(1 1)= -4.88735015E-01 (GPa)
- sigma(2 2)= -4.88735015E-01
- sigma(3 3)= 1.74242410E-01
[Pressure= 2.6774E-01 GPa]

showing the "residue pressure" is about 0.27 GPa,
right?How good is this calculation w.r.t. the residue
pressure? What is the relation between [Pressure=xxx]
and the - sigma(xx) values?

4. I tried to further eliminate the residue pressure,
by increasing strfact to larger value, or decreasing
toldmxf to smaller value, but the residue pressure
seems to change little. Why?

5. The cell relaxation seems to be sensitive to the
input trail lattice parameter. If I use in input file
(a=2.44, c=6.50) or (a=2.46,c=6.70), the final lattice
parameters converge to (a=2.441,c=6.63) and
(a=2.449,c=6.68), respectively. The differences seems
to be too large. Why are the final values sensitive to
input trail values?

6. Generally, how good is cell relaxation?

7. The last question is concerning with the FHI98PP
pseudopotential. Here the potential can be generated
in Harmann type, or Troulier-Martin type. For any
element, do these 2 types of pseudopotential vary in
accuracy? Or they do have same accuracy? Is it true
that in abinit only the Troulier-Martin type
pseudopotentials may be used?

Thank you for reading this.

Linhui


__________________________________________________
Do you Yahoo!?
Yahoo! Platinum - Watch CBS' NCAA March Madness, live on your desktop!
http://platinum.yahoo.com



Archive powered by MHonArc 2.6.16.

Top of Page