This thread is locked.Only browsing is available.
Top Page > Browsing
NEB: how to decide the MD.Opt.criterion value
Date: 2021/01/28 14:24
Name: Borong

In my NEB calculation, I set the MD.Opt.criterion=1.0e-4 (Hartree/bohr) as the default value.
But it can not converge even after more than 560 steps.
I checked the .opt file, and found the minimum |Maximum force| value went down to 0.00086 at step 132,
but it didn't become smaller, after few steps, it went up and finally steady around 0.006, and didn't went down again.

So how can I make it converge?
And can I just set the MD.Opt.criterion=1.0e-3, so that it will converge at step 132 ?
What is the proper way to decide MD.Opt.criterion value?
メンテ
Page: [1]

Re: NEB: how to decide the MD.Opt.criterion value ( No.1 )
Date: 2021/01/30 03:42
Name: Naoya Yamaguchi

Hi,

You can change a part of the path not satisfying the criterion by following
http://www.openmx-square.org/openmx_man3.9/node159.html
And, you may need to monitor the behavior carefully.

Regards,
Naoya Yamaguchi
メンテ
Re: NEB: how to decide the MD.Opt.criterion value ( No.2 )
Date: 2021/02/03 16:09
Name: Borong

Dear Yamaguchi,

Thank you so much for your replay!

Do you mean check the NormRD value in the .DFTSCF file?
As I kept checking the .cif structure file during the NEB calculation,
the structures are normal and reasonable during the calculation from step1 to step560.

And what is the proper way to decide MD.Opt.criterion value?
Do I need to keep the MD.Opt.criterion=1.0e-4 (Hartree/bohr) for every system?

Best regards,
Borong
メンテ
Re: NEB: how to decide the MD.Opt.criterion value ( No.3 )
Date: 2021/02/03 20:25
Name: Naoya Yamaguchi

Dear Borong,

>Do you mean check the NormRD value in the .DFTSCF file?
Yes, it should be checked first.
And, if the initial path is not so bad, you might solve it by increasing the number of images.
I think that because it is the same situation as beads with springs, the more beads (images) are required to describe the more complicated MEP.
And, according to the following,
http://www.openmx-square.org/tech_notes/NEB.pdf
The parallel force tends to make distances in the phase space equivalent to each other so that many beads can draw a smooth curve.
However, I think that you can stop the calculation if the obtained path is good because I'm not sure if strict criterion is needed for unphysical force contributions as mentioned in the above slide.

Regards,
Naoya Yamaguchi
メンテ
Re: NEB: how to decide the MD.Opt.criterion value ( No.4 )
Date: 2021/02/04 19:42
Name: Borong

Dear Yamaguchi,

Thank you so much for your anwser!

By the way,
After getting Geometry optimization of a precursor and a product,
and using the precursor and product structure to do NEB calculation,
I can not get a activation energy after NEB calculation,
the energy of the images kept increasing from image 0 to 6.

Is increasing the number of images the only solution?
Is there other solution for this problem?

Best regards,
Borong
メンテ
Re: NEB: how to decide the MD.Opt.criterion value ( No.5 )
Date: 2021/02/04 20:42
Name: Naoya Yamaguchi

Dear Borong,

>Is increasing the number of images the only solution?
If the precursor and product are correct, you might solve it by giving the initial path manually in the way shown in http://www.openmx-square.org/openmx_man3.9/node159.html
Or, if you know the intermediate image, you can divide the path into two to fix the image, and you can optimize the two paths and combine them.

Regards,
Naoya Yamaguchi
メンテ
Re: NEB: how to decide the MD.Opt.criterion value ( No.6 )
Date: 2021/02/05 10:27
Name: Borong

Dear Yamaguchi,

Thank you so much for your reply!

I will try to change the path to solve the problem.

Best regards,
Borong
メンテ

Page: [1]