[CPMD-list] geometry optimization - isotopes and convergence doubt

[Mark Kosmowski]

On 3/31/08, Axel Kohlmeyer <akohlmey at cmm.chem.upenn.eduAK> wrote:
AK> On Mon, 31 Mar 2008, Mark Kosmowski wrote:
AK>
     MK> Dear CPMD Community:
AK>
AK> dear mark,
AK>
     MK> I am taking the advice given earlier by Axel and running some
     MK> convergence vs. cutoff testing and have a couple questions.  I will
     MK> briefly describe my calculations and then ask my questions.
AK>
AK> that is all good and well, but let me also remind you of the
AK> problem of trying to get an accuracy, that the methods you use
AK> by themselves do not provide. i.d. using DFT, pseudpotentials
AK> and finite (fft compatible) grids.
AK>

Right - our group compares theoretical results to experimental result
to discuss the accuracy of current calculations.

AK> [...]
AK>
     MK> At this cutoff, I found that I needed to use (minimally) 2 k points in
     MK> each dimension to achieve zero negative frequencies, so I am
     MK> performing the convergence vs. cutoff energy using this k point
     MK> condition.  I am now running additional tests at 50, 100, 200 and 250
     MK> Ry to examine geometry and vibrational convergence.  Is this a proper
     MK> strategy?
AK>
AK> yes. however, you are not mentioning the convergence for the geometry
AK> optimization which in turn depends on the (final) convergence for the
AK> wavefunction.
AK>

I am using the default convergence parameters for wave function and
geometry.  (I watch for 10-4 in the geometry steps to know how close
the run is to being done.)

AK>
     MK> At the 150 Ry cutoff, I notice a 0.00017 A bond length difference (out
     MK> of 1.57 A bond length) in the bond of most interest between the 1H and
     MK> 2H geometries.  There is also a 0.000244 (out of -32.65) total energy
     MK> difference.  Is this expected behavior?  If so, do these differences
     MK> have physical meaning?
AK>
AK> not really. with GGA's you have a typical error of about 1% underbinding.
AK> i'd have to see how identical the inputs were. so trying to go far
AK> beyond that is somewhat pointless unless you switch to a different
AK> method (and thus a different code).

The only difference in the inputs was the isotope definition.  I have
a bash script that runs the deuterated case first, then copies to a
new directory and runs a sed script to change the ISOTOPE section from
2H to 1H.  The current run is with light main group elements using LDA
DFT and pade pseudopotentials from the Goedecker directory (downloaded
from the CPMD website).

Once the LDA work is done, I plan on also using at least the PBE
functional and associated pseudopotentials.  Are the pseudopotials of
different functionals different enough to require a full convergence
study by cutoff and k points as I am already doing for LDA / pade?

AK>
     MK> Lastly, at the 50 Ry cutoff for the 2H geometry optimization, I am
     MK> seeing several instances of the following message prior to the final
     MK> geometry:
     MK>
     MK>  !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
     MK>  !! UPDRHO| B2MAX IS VERY HIGH                                 !!
     MK>  !! UPDRHO| DOUBT ON CONVERGENCE. WE TRY ONCE AGAIN            !!
     MK>  !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!
     MK>            1     2.380E-04  -32.291104   0.000E+00   1.00    1.78
     MK>            2     2.216E-04  -32.291104   6.933E-08   6.00    6.22
     MK>
     MK> I am assuming that this just means that 50 Ry is not a high enough
     MK> cutoff energy for my system regardless what the parameter vs. cutoff
AK>
AK> that is more likely because of an instability in the lanzcos
AK> procedure and you'll have to tweak the lanzcos related parameters
AK> to get it to run more stable. it looks also like the code detected
AK> that and resetted the diagonalization.

Thank you for explaining this.

AK>
AK> cheers,
AK>   axel.
AK>
     MK> plots look like.  Is this a correct conclusion?
     MK>
     MK> Thank you,
     MK>
     MK> Mark Kosmowski
     MK>
     MK> Graduate Student
     MK> Physical Chemistry
     MK> Syracuse University
     MK> Syracuse, NY, United States
     MK> _______________________________________________
     MK> CPMD-list mailing list
     MK> CPMD-list at cpmd.org
     MK> http://cpmd.org/mailman/listinfo/cpmd-list
     MK>
AK>
AK> --
AK> =======================================================================
AK> Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
AK>   Center for Molecular Modeling   --   University of Pennsylvania
AK> Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
AK> tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
AK> =======================================================================
AK> If you make something idiot-proof, the universe creates a better idiot.
AK>