[CP2K-user] [CP2K:21950] Re: Different eigenvalues from Hybrids + ADMM calculations in CP2K 2025.1 compared to 2024.1

[Max Pierre]

That was indeed the issue! For some reason in cp2k-2024.1 the ADMM band 
gaps were very close to the normal PBE0 ones even with purification 
switched on, so I hadn't noticed that anything was off before switching to 
2025.1. I guess I was just lucky before that using the wrong purification 
setting wasn't causing any big issues for me in 2024.1.

Thank you for the help!
Max

On Wednesday, 5 November 2025 at 14:12:28 UTC+1 fabia... at gmail.com wrote:

> Hi Max,
>
> I don't know where the difference comes from. But to get the band gap you 
> need to switch off purification, so neither of those values are reliable. 
> Change  ADMM_PURIFICATION_METHOD MO_DIAG to ADMM_PURIFICATION_METHOD NONE.
> See, e.g. here: 
> https://groups.google.com/g/cp2k/c/WAdLCCMk47A/m/fP0AeCaRBAAJ
>
> Cheers,
> Fabian
> On Wednesday, 5 November 2025 at 10:54:38 UTC+1 Max Pierre wrote:
>
>> Hello CP2K community,
>>
>> I have recently switched from cp2k-2024.1 to cp2k-2025.1 and initially 
>> got some strange output. Running some test calculations using the same 
>> input with both versions, I found that the energy eigenvalues are shifted, 
>> and the HOMO - LUMO gap is much larger in 2025.1.
>>
>> - 2024.1:
>> *
>>       -0.15118136     -0.15058927     -0.15030345     -0.14919531
>>       -0.14827092     -0.14449236     -0.14178774     -0.14087384
>>       -0.14005256
>>  Fermi Energy [eV] :   -3.811024
>>   
>>   Lowest Eigenvalues of the unoccupied subspace spin            1
>>  -----------------------------------------------------
>>  OT| Eigensolver reached convergence in 763 iterations
>>        0.19181627
>>   
>>  HOMO - LUMO gap [eV] :    9.030610
>>
>>  ENERGY| Total FORCE_EVAL ( QS ) energy [a.u.]:           
>>  -1322.383368758011557
>> *
>>
>> - 2025.1:
>> *
>>       -0.25036349     -0.25011266     -0.24923456     -0.24855820
>>       -0.24640841     -0.24154336     -0.23787275     -0.23769990
>>       -0.23629481
>>  Fermi Energy [eV] :   -6.429909
>>   
>>   Lowest Eigenvalues of the unoccupied subspace spin            1
>>  -----------------------------------------------------
>>  OT| Eigensolver reached convergence in 887 iterations
>>        0.25231706
>>   
>>  HOMO - LUMO gap [eV] :   13.295805
>>
>>  ENERGY| Total FORCE_EVAL ( QS ) energy [hartree]         
>>  -1322.383368757196422
>> *
>>
>> Both calculations use PBE0 with ADMM pFIT3 and are started from a RESTART 
>> file from a PBE calculations, the system is amorphous Polyethylene. The 
>> 2024.1 HOMO-LUMO gap was pretty reasonable, the newer one is very high 
>> compared with experimental values and previously reported hybrid functional 
>> results. 
>>
>> I didn't find anyone else experiencing this issue, does anyone have any 
>> idea what might be going on? 
>>
>> Thank you,
>> Max
>>
>>

-- 
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/e9584487-14f4-4fb5-a760-905176311dd0n%40googlegroups.com.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20251106/9dc39049/attachment-0001.htm>