[CPMD-list] General procedure to start a NPT CPMD simulation

[Axel Kohlmeyer]

On Fri, 28 Jul 2006, junwang wrote:

> Dear All:
> 
> Thanks for your timely reply. I began to understand some mechanism of a NPT CPMD simulation.
> 
> However, I am still interested to know the general procedure to start a NPT CPMD simulation. 
> 
> 1. Of course, the first step is always to do a wavefunction optimization.

> 2. After that I am not sure whether we should start PARRINELLO-RAHMAN
> NPT immediately [Procedure 1] or control the system temperature first
> (including TEMPCONTROL IONS step,NOSE IONS step and NOSE ELECTRONS
> step) followed by pressure control(initiate PARRINELLO-RAHMAN
> NPT)[Procedure 2]?

well, if you look below, your wavefunction is desintegrating after
the first step. so i recommend you start first by equilibrating
your system and making sure that you have parameters, that result
in a stable CP dynamics. later you can start a variable cell dynamics.

your input however has a few problems:

for variable cell you should provide a reference cell
dimension that will be the largest possible cell for 
your dynamics. your setup only works, if you start from
an expanded system and then crank up the pressure.

your plane wave cutoff is far too low. you have to make
sure, that your stress tensor is converged wrt. to the
basis set, which does not happen as fast as the forces,
even for an 'ideal' element like silicon. so even if
you get good forces (=structures with ~12 ry, you will
need much more 40-50ry for the stress tensor).

on top of that, you have to consided, that CPMD does
a calculation with a constant number of plane waves,
i.e. your _effectice_ basis set size will change
when the box compresses or expands resulting in 
so-called pulay stress. in other words, you have 
to make perfectly sure that your stress tensor is
_very_ well converged wrt to the plane wave cutoff.

as you may see now, doing variable cell calculations
with a plane wave basis set is not a trivial as
in classical MD. so perhaps it would be useful
if you give us some hints of want you want to
use those calculations for. 
there may be alternative and more convenient ways
to achieve your goal.

cheers,
   axel. 


[...]

> 
> However, even if I increased SPLINE RANGE to 5.0, the warning" SPLINE REGION SMALLER THAN MAXIMUM G-VALUE" is still reported in the output file as shown below:
> 
>     NFI    EKINC    EKINH  TEMPP           EKS      ECLASSIC          EHAM         DIS     TCPU
>          1  1.28022      NaN    NaN     -84.72569           NaN           NaN   0.360E+10     0.09
>  WARNING| SPLINE REGION SMALLER THAN MAXIMUM G-VALUE
>          2      NaN      NaN    NaN           NaN           NaN           NaN         NaN     0.08
>  WARNING| SPLINE REGION SMALLER THAN MAXIMUM G-VALUE
>          3      NaN      NaN    NaN           NaN           NaN           NaN         NaN     0.09

> What is the possible reason for this? Besides that some energy values
> are shown as "NaN", I don't know why?

your system is 'exploding'.


> Output file also contains the following warnings:
> 
>     NFI    EKINC    EKINH  TEMPP           EKS      ECLASSIC          EHAM         DIS     TCPU
>          1  8.76383  0.00000 1500.0     100.83262     100.93949     109.70332   0.652E-05     0.29

... here it is even worse than before. compare EKINC to typical values
for 'normal' CP dynamics.



> Best regards,
> Sincerely,
> Jun WANG
> 
> Department of Chemistry
> University of Nebraska-Lincoln
> 536 Hamilton Hall
> Lincoln, NE 68588-0304
> U.S.A.
> E-mail:junwang at unlserve.unl.edu
>        junwang at bigred.unl.edu
> 
> 
> _______________________________________________
> CPMD-list mailing list
> CPMD-list at cpmd.org
> http://cpmd.org/mailman/listinfo/cpmd-list
> 

-- 
=======================================================================
Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
   Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
=======================================================================
If you make something idiot-proof, the universe creates a better idiot.