Energy Fluctuations Without Convergence

[timw...@hotmail.com]

Dear CP2K Users,

I'm running a geometry optimization for Ni(111) containing 288 atoms. However, the energy seems to fluctuate a lot, and the system does not tend to converge. Could someone give me some advice on what I can do in this case? My input and output files are attached below. Thanks!

Best regards,
Tim

[Thomas Kühne]

Dear Tim, 

your alpha value is definitely too large, try 0.1 or smaller. 

Best, 

Thomas

--
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/91bc2d9b-5559-4e92-ab6f-4be3f2225393n%40googlegroups.com.
<input.xyz><Ni_111-pos-1.xyz><input.cp2k><output.cp2k>

==============================

Thomas D. Kühne

Dynamics of Condensed Matter

Chair of Theoretical Chemistry

University of Paderborn

Warburger Str. 100

D-33098 Paderborn

Germany

thomas...@upb.de

+49/(0)5251/60-5726

[timw...@hotmail.com]

Dear Thomas,

Thanks for your suggestion! I changed the alpha value to 0.1, and the energy did become less fluctuating. However, the status in the convergence check still showed NO for all four parameters after 21 optimization steps (as shown below). Does it mean that I need to lower the alpha value further? Or will it become YES in the end even if I don't change the alpha value anymore?

Best regards,

Tim

[Thomas Kühne]

Dear Tim, 

this is to say that your SCF to compute the electronic structure incl. total energy and most importantly the forces is converging. 

Your geometry optimization loop, however, has just completed 21 iterations and is not converged as yet. For CG and BFGS all 

criteria must be YES in order to complete your geometry optimization. However, LBFGS may be fast converging, but I am not 

sure if they rely on the same criteria and not on the WANTED* criteria. 

Best, 

Thomas

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/f4b636f1-8f1d-466a-a750-6aa7a2821270n%40googlegroups.com.
<Convergence Check.PNG>

[Tim Wang]

Dear Thomas,

Thanks again! It finally converged after 64 optimization steps with all the criteria being YES.

Best regards,

Xiao Wang

---

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Thomas Kühne <tku...@gmail.com>
Sent: Thursday, January 20, 2022 9:01 PM
To: 'Dorothea Golze' via cp2k <cp...@googlegroups.com>
Subject: Re: [CP2K:16474] Energy Fluctuations Without Convergence

 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/50F8703F-A7FC-468D-916C-B0A0DADF3933%40gmail.com.

[Yike Ye]

Hi Tim,

Did your scf converge? I also calculated Ni(111) slab and it can converge by changing the criteria but the scf didn't converge after even 500 steps.

Best Regards

Yike

[Tim Wang]

Hi Yike,

For me, the scf did converge.

Best,

Tim

---

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Yike Ye <yike...@gmail.com>
Sent: Monday, April 11, 2022 11:55 PM
To: cp2k <cp...@googlegroups.com>
Subject: Re: [CP2K:16837] Energy Fluctuations Without Convergence

 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/51840dc8-82b5-4d5d-8fb1-3cc0238e5ee1n%40googlegroups.com.

[Yike Ye]

Hi Tim,

Is it possible for you to put your input and output on this chat? I tried to change α to 0.04 but it didn't work. My initial guess is atomic rather than restart. Also, I don't think the multiplicity in your input file should be set to 1 because that means on spin electron. 

Best Regards

Yike

[Tim Wang]

Hi Yike,

I attached my input and output files below. I set the initial guess to restart because for my simulation, it has 288 atoms, and the job cannot complete in the time I requested. For the first run, I think atomic and restart are the same since in the CP2K manual, it says "ATOMIC if not present". For the multiplicity, I just did a preliminary test to see if CP2K can handle this large system or not (my friend cannot get convergence using VASP), but I agree that the multiplicity should be further optimized.

Best,

Tim

---

From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Yike Ye <yike...@gmail.com>

Sent: Thursday, April 14, 2022 5:37 PM
To: cp2k <cp...@googlegroups.com>
Subject: Re: [CP2K:16855] Energy Fluctuations Without Convergence

 

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/b3c75404-58fe-440e-acbc-fe2321dac77dn%40googlegroups.com.