[CP2K-user] [CP2K:22187] Re: SCF converges more slowly as AIMD progresses, but resets when restarting

[Yinyu chan]

Hi,

Thank you for pointing this out. You are correct that a clear energy jump 
occurs at step 5001 where I restart the simulation as a trial to speed up 
convergence with a RESTART file and modify the ASPC order and NDIIS to 2 
and 5, respectively. The settings keeps consistent in the whole production 
run stage. However, the issue raised in this conversation still exists.

Regards,
Yinyu
在2026年4月15日星期三 UTC+8 17:14:14<Marcella Iannuzzi> 写道：

> Hi 
>
> you have a bigger problem than the time to convergence. 
> At restart there is a clear jump in energy, indicating that the dynamics 
> is restarting from a different state.
> Regards
> Marcella
>
> On Wednesday, April 15, 2026 at 10:55:10 AM UTC+2 mayso... at gmail.com 
> wrote:
>
>> Hi,
>>
>> It doesn't seem to work. I wonder if restarting the AIMD every 
>> certain number of steps yields the normal results. If so, simulations can 
>> be accelerated in this manner.
>>
>> 在2026年4月13日星期一 UTC+8 20:53:24<Shiyang Wang (W0728)> 写道：
>>
>>> You can consider using an alternative preconditioner, such as FULL_ALL, 
>>> and see how it goes.
>>> 在2026年4月13日星期一 UTC+8 15:42:44<Yinyu chan> 写道：
>>>
>>>> Hi,
>>>>
>>>> Thanks for your reply!
>>>>
>>>> Actually, the input settings follow a JACS Au paper on the Au/NaCl 
>>>> interface without smearing. And OT method is required for applying periodic 
>>>> external electric fields. I suspect the slow SCF convergence is related to 
>>>> the use of historical information in cp2k, but i haven't found any useful 
>>>> info on it so far.
>>>>
>>>> Best regards,
>>>> Yinyu
>>>> 在2026年4月10日星期五 UTC+8 15:44:09<Shiyang Wang (W0728)> 写道：
>>>>
>>>>> Hi,
>>>>>
>>>>> I have doubt if your system is fit for OT method; it looks like to be 
>>>>> metal-based and may require smearing.
>>>>>
>>>>> 在2026年4月9日星期四 UTC+8 19:18:49<Yinyu chan> 写道：
>>>>>
>>>>>> Dear CP2K community,
>>>>>>
>>>>>> I'am a new user of CP2K(v2026.1) and I'm running an AIMD simulation 
>>>>>> of a metal-electrolyte interface (Au-Pd alloy/NaCl(aq)) with electric field.
>>>>>>
>>>>>> However, I encountered an issue where the SCF starts fast but then 
>>>>>> becomes slow, as shown in the *ener file. But when I restarted the AIMD at 
>>>>>> steps 5000(SCF_GUESS RESTART) and 5378(SCF_GUESS ATOMIC), the SCF 
>>>>>> convergence starts fast again, but then becomes slow once more. I would 
>>>>>> like to know whether this phenomenon is normal, and how to keep the SCF 
>>>>>> convergence fast during the AIMD. 
>>>>>>
>>>>>> I've attached the inputs and outputs for reference. I would 
>>>>>> appreciate it if anyone could offer suggestions on this issue.
>>>>>>
>>>>>> Thank you for your time and support.
>>>>>>
>>>>>> Best regards,
>>>>>> Yinyu Chan
>>>>>>
>>>>>

-- 
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+unsubscribe at googlegroups.com.
To view this discussion visit https://groups.google.com/d/msgid/cp2k/c4366eb1-ee9c-4336-a5e2-f615aeabb1b3n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20260416/a6e08ad1/attachment-0001.htm>