[Wien] difference of min- energy point and zero -force point

[naser zare]

Dear Stefaan,
  Dear Laurence,
   
  Thank you very much for your useful and extended replies. I was just repeating my pseudo-potential calculations by this code by taking the number of k-points as minimum as possible. Now, by your comments I learn that these two minimization methods would agree (within at most 0.1 mRy/au) only for sufficiently large number of k points.
  Thank you again for clarifying this point.
   
   Best Regards.
 Nasser Zare Dehnavi

Stefaan Cottenier <Stefaan.Cottenier at fys.kuleuven.be> wrote:
  
> 3)I use RK=7 and mesh 18X18X18 for bulk and 10X10X1 for 
> asupercell with 38 atoms and a surface cell(2X2) .I study about 
> work function and I use from this parameters because :a)give me good 
> value for workfunction and b)for avoid of computational cost 
> of large parameters.

Don't take this personally, but these are very bad and even dangerous 
arguments. As computational physicists, we MUST NOT take the 
parameters that give us good agreement with experiment. We must take 
the parameters that give us numerically accurate results within the 
framework we work in (e.g. DFT-LDA). If we don't do that, we're just 
doing meaningless fitting, not physics. If it turns out you find the 
computational cost is too much for the quantity you want to study, 
then so be it. Put the problem aside and do something else instead of 
doing something meaningless. In your case, this means you must verify 
whether going to a larger RKmax or fine k-mesh does not change you 
work function to an amount that would be large enough to alter your 
physical conclusions. As long as you are not sure about this, there is 
no point in proceeding with the rest of your work.

There is a good reason why I spent some time writing this down: a bad 
approach to DFT studies affects the entire community. Imagine you come 
to some conclusion based on an unjustified small basis set. Perhaps 
you'll even make a prediction based on your results. Then an 
experimentalist tries to verify your prediction, and find it to be 
wrong. Obviously, the conclusion of his or her paper will be: DFT 
gives bad predictions. If this happens too often, then DFT gets an 
unjustified bad image, and few people will believe DFT predictions in 
the future, even if they are made with good basis sets. Our entire 
community has a responsibility in keeping up the credibility of DFT: 
there are enough places where it doesn't agree with experiment, we 
should not create additional ones by being too tolerant on numerical 
issues as basis set and k-mesh.

I hope you seriously think about this before continuing!

Now back to your problem:

As you apparently did not test the accuracy of your forces with 
respect to k-mesh and basis set, it might be that you are working with 
too much numerical noise on the forces, leading mini to an incorrect 
result. So test the accuracy of your forces.

This is an interesting issue:

> with attention to this point that min-energy is obtained in 
> smaller time by plot total energy versus distance of atom from 
> surafce with respect to use of mini..

As the force is the gradient of the total energy surface, this 
statement cannot be true. Even if you don't use mini and displace the 
atoms yourself, it is worthwhile to calculate the force in the last 
iteration and plot not :ENE but the force as a function of 
displacement. It will show you much better in which direction and by 
which amount you have to displace, requiring less steps to find the 
equilibrium than if you use only the energy information. As a check, 
of course, you can look at the energies afterwards. If your numerical 
precision was OK, the zero of the force and the energy minimum must 
agree.

Stefaan


Quoting naser zare :

> Dear Stefaan,
> Thank you very much for your reply.
> I should mention that:
> 1)the discrepancy in position is 0.122A(1.544A for min-energy and 
> 1.666 A for zero-force)
> 2)convergence criteria for energy is 0.0001ry and for charge is 
> 0.002 and when I use from
> mini for force is 4 mry/au
> 3)I use RK=7 and mesh 18X18X18 for bulk and 10X10X1 for 
> asupercell with 38 atoms and a surface cell(2X2) .I study about 
> work function and I use from this parameters because :a)give me good 
> value for workfunction and b)for avoid of computational cost 
> of large parameters.
> 4)when I used mini , I move only one atom ( like when I obtain 
> min of energy ) and all atoms of slab are fixed.
>
> I have a question,can I fell sure that point obtained from plot 
> of total energy versus distance of atom from surafce is true 
> point and my problem is calculation of force?with attention to this 
> point that min-energy is obtained in smaller time by plot total 
> energy versus distance of atom from surafce with respect to use 
> of mini..
>
> thank you very much for your attention.
> Best Regards.
> Nasser Zare Dehnavi
>
>
> Stefaan Cottenier wrote:
>
> You need to provide more information, for instance:
>
> * How big is the discrepancy in position?
> * How well are your calculations converged and what are your
> convergence criteria?
> * Did you verify your basis set and k-mesh are well enough to provide
> good forces?
> * When you used mini, did you move only one atom, or all of them?
> *...
>
> Stefaan
>
>
>
> Quoting naser zare :
>
>> Hello
>> Dear all:
>> I use from Wien2k 6.4. I want to obtain position of adsorption of
>> atom on a surface.I do it in two way :1)I move atom towards
>> surface step by step and I plot total energy
>> versus distance from surface and I obtain point of min energy and
>> 2)I relax position of atom by MINI, but these two point are not
>> the same.on the other hand ,point that obtained by Mini has not
>> lowest energy. I want to know reason of this problem .thank you
>> very much for your attention.
>> Best Regards.
>> Nasser Zare Dehnavi
>>
>>
>>
>>
>> ---------------------------------
>> Sucker-punch spam with award-winning protection.
>> Try the free Yahoo! Mail Beta.
>
>
>
> --
> Stefaan Cottenier
> Instituut voor Kern- en Stralingsfysica
> K.U.Leuven
> Celestijnenlaan 200 D
> B-3001 Leuven (Belgium)
>
> tel: + 32 16 32 71 45
> fax: + 32 16 32 79 85
> e-mail: stefaan.cottenier at fys.kuleuven.be
>
>
> Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>
>
>
> ---------------------------------
> It's here! Your new message!
> Get new email alerts with the free Yahoo! Toolbar.



-- 
Stefaan Cottenier
Instituut voor Kern- en Stralingsfysica
K.U.Leuven
Celestijnenlaan 200 D
B-3001 Leuven (Belgium)

tel: + 32 16 32 71 45
fax: + 32 16 32 79 85
e-mail: stefaan.cottenier at fys.kuleuven.be


Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm

_______________________________________________
Wien mailing list
Wien at zeus.theochem.tuwien.ac.at
http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien


       
---------------------------------
TV dinner still cooling?
Check out "Tonight's Picks" on Yahoo! TV.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20070411/95c036d7/attachment.html