metal ions in cp2k

[Qiang Li]

Dear all,

I found several publications dealing with metal ions in the AIMD calculations, Zn(2+) for example. The way I see it is that the authors use Cl- counter ions to neutralize the system. However, how one can make sure that the charge of the two ions are well separated, i.e. Zn (2+) and Cl (-1) ? is there any charge analysis available  in CP2K 4.1 ? Any suggestions are appreciated.

All the bests,

Li

 

[Vladimir Rybkin]

Dear Li,

there are Mulliken and Hirshfeld population analyses in CP2K, which provide partial charges on atoms.
To keep charges separated you need a large simulation cell. Alternatively, you may consider a charged cell (no counter ions). This is simpler to do, although, the large cell is even more a must.

Yours,

Vladimir

вторник, 8 ноября 2016 г., 15:31:07 UTC+1 пользователь Qiang Li написал:

[Qiang Li]

Hi, Vladimir,

Thanks for your reply, that was quite helpful.

I did some tests before I reply back. Simply, I calculated Na+Cl- system (distance is set to 10 angstrom, using a 20x10x10 box), and did the Mulliken charge analysis. I found that using  GPW method, i.e. using the GTH potential and DZVP basis set (same method in this paper: http://dx.doi.org/10.1088/2053-1583/3/2/025026), the charges was Na(0.5)Cl(-0.5), which is very strange to me. Alternatively, the charges are well separated when I apply GAPW method (Na+Cl-).

What's the physics behind this?

Best regards,

Li

在 2016年11月8日星期二 UTC+8下午11:03:55，Vladimir Rybkin写道：

[Vladimir Rybkin]

Hi Li,

this means you have a "covalent" dissociation into atoms, not ions. Try to do a spin-unrestricted calculation, i.e. UKS (alternative name - LSD) calculation in the DFT section.

Yours,

Vladimir

четверг, 10 ноября 2016 г., 12:01:16 UTC+1 пользователь Qiang Li написал:

[Qiang Li]

Dear Vladimir,

The UKS does not help to separate the charges, as I expected. I did a calculation with same system by Gausssian 09 for a test, and the Mulliken charge analysis shows that the Na(+1)Cl(-1) are well separated. I do expect similar result from CP2k, since it also describes the valence electrons with gaussian basis sets.  I am uploading the input file for my cp2k run, which you can have a quick look whether I did something wrong or not.

Thank you for being replying.

Best regards,

Qiang

在 2016年11月10日星期四 UTC+8下午7:51:26，Vladimir Rybkin写道：

[Matt W]

Hi Qiang,

I think that dissociation into partially charged ions is the 'correct' incorrect behaviour using standard functionals - see

'@article{ruzsinszky2006spurious,

  title={Spurious fractional charge on dissociated atoms: Pervasive and resilient self-interaction error of common density functionals},

  author={Ruzsinszky, Adrienn and Perdew, John P and Csonka, G{\'a}bor I and Vydrov, Oleg A and Scuseria, Gustavo E},

  journal={The Journal of chemical physics},

  volume={125},

  number={19},

  pages={194112},

  year={2006},

  publisher={AIP Publishing}

}'

for instance. 

Also, it is quite hard to converge wavefunctions for the stretched molecule - I'm not sure your set up really converges. I suspect that Gaussian converges to a different solution - though you don't say what setup you are using there.

So, actually what you are trying to do is quite difficult, rather than an easy task. A better approach is to model the NaCl molecule near equilibrium then calculate the Na and Cl atoms separately for most thermodynamic applications.

Matt