[CPMD-list] help on the kohn-sham energies!

[Axel Kohlmeyer]

On Tue, 13 Dec 2005, [gb2312] Shangyi Ma wrote:

> Dear axel and cpmd-users:
>    Thanks for your kindly reply!
>    Today I have other questions need your help!
>    Firstly, I have optimized the wavefunction of the 5 atoms clusters and recalculated the kohn-sham energyes of it, but I find the final total energys is different,but I don't know what's the reason cause the difference. The following are the input and out files in different calculation respectively:
>
> input file for the wavefunction optimization:
> *********************************************
> &CPMD
>  OPTIMIZE WAVEFUNCTION
>  CONVERGENCE ORBITALS
>  1.0D-5
>  FREE ENERGY FUNCTIONAL
>  LANCZOS DIAGONALISATION

if you are doing a diagonalization to optimize the
wavefunction, there is little need to do a kohn-sham
energies calculation. you already have the diagonalized
orbitals.

>  LANCZOS PARAMETAERS N=5
>  1  6  20  1.0D-9
>  0.05  1.0D-10
>  0.01  1.0D-12
>  0.0025  1.0D-14
>  0.001  1.0D-16
>  ANDERSON MIXING

> input for the khon-sham energies:
> ********************************
> &CPMD
>  KOHN-SHAM ENERGIES
>  8

do you really need 8 _additional_ empty states.
the majority of them will be meaningless.

>
> You can see the difference what I mean from the above files. The final 
> results "TOTAL ENERGY"  "EIGENVALUES(EV) AND OCCUPATION" and "CHEMICAL 
> POTENTIAL(is it the Fermi energy?)" are all not the same for the 
> wavefunction calculation and kohn-sham energies calculation. So what 
> cause the difference and which results are more correct? Are there 
> some faults in my 2 input files?

>  Secondly, what's the meaning of "NC" in the "EIGENVALUES(EV) AND 
> OCCUPATION" results if the results show the occupation stats? Does it 
> mean "not converged"? Is there any difference between the "13

EXACTLY. so the KS energies run added one more diagonalization
iteration to the not fully diagonal wavefunction, a few more
got converged now. so you have to continue the wavefunction optimization
for a bit, or change the lanzcos parameters, so you have a few
more lanczos iterations before you update the wavefunction, not only
one.

regards,
      axel.

> -1.3219857 0.000 " and "14 -0.8650114NC 0.000"? The occupation of all 
> them are 0 and "14" indentfied NC and "13" is not.

>
>  Any comments and advices are appreciated!
>  Thanks in advance!
>  Best regards!
>  shyma
>

=======================================================================
Axel Kohlmeyer    e-mail: akohlmey at cmm.upenn.edu, tel: ++1-215-898-1582
    Center for Molecular Modeling    --    University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.