Energy calculation for charged system LSD/UKS ?

[Andreas Döll]

Dear cp2k community,

I have two charged systems I want to perform a single energy calculation.

System a) Neutral molecule with a Lithium at some distance

System b) Charged molecule (cut out of a larger molecular structure, so 3 Hydrogens are missing) 

For both calculations, my energy either diverges or makes jumps. 
Usually I have performed calculations with neutral molecules, not charged systems, so I am not familiar with the proper procedure. I followed the example give here https://www.cp2k.org/exercises:2014_ethz_mmm:hfx_h2ion 

I defined my Charge in the DFT section, as well as using LSD option and switched to a Nonperiodic system - as far as I understand this is necessary? Do I need to define the multiplicity of my system as well ?

Cheers, 

Andreas

[Andreas Döll]

Hello, 

I just want to update people who may have the same question in the future. After further research, my problems could be partially solved with the following changes to my input script: 

...

CHARGE +1
    LSD
    MULTIPLICITY 1

...

POISSON_SOLVER

PERIODIC NONE

...

&CENTER_COORDINATES
        &END
...

All of this, and a sufficiently large cutoff (my research shows 5+ A) worked for me for case a) where I have a multiplicity of 1. Centering is there to prevent problems at the edge of the box with the solver.

Sadly for system b) where the multiplicity is higher, I always get the usual wavelet solver issue, even though I 

a) Center my atoms

b) Have a large box (more than 10 A from edges even)

c) have a large cutoff

I keep getting problems with the Wavelet solver at the edge: Usually, the error message

 *** WARNING in ps_wavelet_methods.F:238 :: Density non-zero on the edges ***
 *** of the unit cell: wrong results in WAVELET solver                    ***

can be treated with a larger box, it seems to me from looking around the Cp2kgroup that higher multiplicity might be difficult, but even when increasing the box to have 20+ A vacuum in each direction I still get a problem at the edge. I attach my input script for this issue, maybe somebody can spot some obvious mistake quickly! Is this maybe a chemistry issue (getting multiplicity wrong? I just cut out a molecule off a larger structure, so at each branch of the Triazine, a Hydrogen is missing)

Thank you for your patience

[bmk]

It could be related to your EPS_SCF and EPS_DEFAULT, which are both quite high. For DFT-MD I usually set EPS_SCF to 1.0E-6 and EPS_DEFAULT to 1.0E-12, and for accurate single points I set EPS_SCF to 1.0E-8 and EPS_DEFAULT to 1.0E-16. Check https://pubs.acs.org/doi/full/10.1021/acs.jpcb.5c05851 section 2.1.4 for more details.

The amount of OT inner SCF cycles also seems quite high to me.

Op donderdag 5 februari 2026 om 17:07:27 UTC+1 schreef Andreas Döll:

[Andreas Döll]

Thank You for Your answer. 

I know about the new "tutorial paper" and would recommend anyone to read it. But why would EPS be the cause of my problem? This cannot be the cause for density at the edges? These values were for testing anyways, for production I make my convergence criterion more tight.

Changing the following things does nothing:

CUTOFF 1000 

EPS_DEFAULT 1.0E-9

EPS_SCF 1.0E-8

I played around and still, immediately at step 1, I get density at the edge error message. I dont think EPS values are the source of the problem

Any input is appreciated!

Cheers, 

Andreas