About the Dirichlet boundary

[Lakshmi Anil]

Dear All,

I am new to CP2K. I was trying to give an electric field to my Ru metal surface(144 atoms)with adsorbed nitrogen. For applying the electric field, I used an Implicit Poisson solver and Dirichlet boundary condition are used. Unfortunately, I got an error that the Dirichlet extents are outside the simulation cell even though I gave the same simulation cell dimension. Then, I reduced it to half to check whether it works, but it again gave me the same error. Could anyone please help me with that? I am attaching the input file as well as the output file for the reference.

Thanks,

Lakshmi.

[Marcella Iannuzzi]

Dear Lakshmi

I suspect that the problem might come from the intercept 25, but I do not have experience with the method.

Regards

Marcella

[Lakshmi Anil]

Dear Marcella,

I was trying to repeat the same system given in the article :

DOI

https://doi.org/10.1039/D2CP05052H. So in this, they are using the distance between the upper and lower surface as 25A0. I  try to do it in the same way. But the same error is shown. Do you have any suggestions?

Thanks,
Lakshmi

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/9eca8e1d-6a1d-47de-b1cb-457a81ad05e7n%40googlegroups.com.

[Marcella Iannuzzi]

I see, 

but which box size do they use in z? you have 21.7 which is smaller than 24, and a thickness of 0.5. 

Have you tried with a larger box?

Best

Marcella

[Lakshmi Anil]

The size along z is not mentioned in the article. They are using around 200 atoms structure. I will try by using a smaller intercept value and also one with a larger box.

Thank you so much.

Lakshmi.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/44d4d72c-8a2c-40a8-8fa0-004d88f7baacn%40googlegroups.com.

[Lakshmi Anil]

Dear All,

I had tried with a large box as well as decreasing the intercept and thickness. The simulation started running but it was not converging. It almost ran for 10 hours but it didn't converge. I don't understand what is missing in my input. Could anyone please help me to solve the problem? I am attaching both the edited input and the output I got for your reference.

Thanks,

Lakshmi.

[Marcella Iannuzzi]

Hi Lakshmi

The posted output file has data from two un-finished runs. The first part seems to correspond to a run where the convergence was not too bad.

Why was it stopped? The second seems not to be restarted from the previous one.

In general the Broyden mixing works better with a smaller ALPHA.

Have you tried to use an extrapolation method? 

Regards

Marcella

[Lakshmi Anil]

Hi Marcella,

I have set the time for the simulation to 10 hours. I am not aware of how much time it takes to complete the simulation. I never tried the extrapolation method. Can you please explain to me how the extrapolation method is applied?

Thanks,

Lakshmi.

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/1a9af2ac-e679-429f-9b56-bd256053b32an%40googlegroups.com.

[Marcella Iannuzzi]

Hi Lakshmi

If you set a job time, then it is  good practice to also give a proper 

• WALLTIME

  in input, such that cp2k can close up the calculation before being killed by the scheduler. This is the best way to get proper restart files such that one can restart the simulation if the allocated time was not sufficient.  

  
  

  You did not explain why the output you posted contains data from two different simulations, one that seems to converge better than the other. 

  One part of the output you posted shows a geo_opt running over 12 iterations, the gradients are not huge, but still high. I wonder whether you tried to run geo_opt for the same system without the complications of the implicit poisson solver first. Such an optimization can take many hours, depending on many different factors 
  

  
  

  
  

  The extrapolation method is selected with 

• FORCE_EVAL
 
• DFT
 
• QS EXTRAPOLATION

• there are different choices, which are better then just the atomic guess.

Why are you activating the kpoints if you then require only gamma? By doing this, there are several restrictions to the available features. 

Regards

Marcella