How to use ECP potential in FORCE

Eugene

unread,

Feb 7, 2022, 11:30:11 AM2/7/22

to cp2k

According to cp2k documentation https://manual.cp2k.org/trunk/CP2K_INPUT/FORCE_EVAL/SUBSYS/KIND.html#POTENTIAL it is possible to set POTENTIAL variable value as ECP, in order to use Effective core potential. I've converted basis set from http://www.tc.uni-koeln.de/PP/clickpse.en.html according to CP2K documentation. The section KIND in my input file is:

&KIND V
BASIS_SET STOLL-PBE
POTENTIAL ECP
&POTENTIAL
V nelec 10
V ul
2 1.000000000 0.000000000
V S
2 14.490000000 178.447971000
2 6.524000000 19.831375000
V P
2 14.300000000 109.529763000
2 6.021000000 12.570310000
V D
2 17.480000000 -19.219657000
2 5.709000000 -0.642775000
&END
&END KIND

The basis set in new file is:

V STOLL-PBE
4
2 0 0 7 1 0 # Gauss type functions l=0 set 1 (3s orbital)
12.8432080 0.460448
11.3757530 -0.495309
5.4069740 -0.350821
1.4659270 0.654979
0.5980800 0.582548
0.0887900 0.017081
0.0353180 -0.007455
2 0 0 7 1 0 # Gauss type functions l=0 set 2 (4s orbital)
12.8432080 0.126932
11.3757530 -0.139770
5.4069740 -0.089862
1.4659270 0.204669
0.5980800 0.321169
0.0887900 -0.657043
0.0353180 -0.474893
2 1 1 6 0 1 # Gauss type functions l=1 set 1 (3p orbital)
31.8898680 0.004658
8.2371780 -0.119249
4.3283730 0.125195
1.5405260 0.620704
0.5280810 0.416780
0.0899620 0.014015
2 2 2 5 1 # Gauss type functions l=2 set 1 (3d orbital)
22.6804330 0.031628
6.8613120 0.153807
2.2754450 0.378661
0.7319220 0.507276
0.2007460 0.275921

CP2K starts normally but exit with error message:

CPASSERT failed qs_mo_types.F:295

By checking source code I have error here:

CPASSERT(nmo >= nomo)

Could someone kindly give a working example for force calculations with ECP potential.

Many thx in advance,
Eugene

Eugene

unread,

Feb 7, 2022, 1:42:12 PM2/7/22

to cp2k

OK, here is Update. Actually I'm testing ECP with V2O5 crystall 1x3x2 Supercell. Now when I switch to GTH pseudopotential I have the following initial values for number of MO:

Total number of - Atomic kinds: 2
- Atoms: 84
- Shell sets: 84
- Shells: 684
There are 24 V atoms and 60 oxygens. I.e. total number of shells is: (24*13+60*6)=672 (for V there is 13 valence electrons for O it's 6). Which is different then 684 for some reason.

In case of ECP the numbers are:

Total number of - Atomic kinds: 2
- Atoms: 84
- Shell sets: 336
- Shells: 192

Well, Huston we have a problem. How to resolve the issue?

понедельник, 7 февраля 2022 г. в 19:30:11 UTC+3, Eugene:

hut...@chem.uzh.ch

unread,

Feb 7, 2022, 1:54:50 PM2/7/22

to cp...@googlegroups.com

Hi

check your basis set definition: the p shell says:

2 1 1 6 0 1 # Gauss type functions l=1 set 1 (3p orbital)

That would be zero p-orbitals. Probably a typo. It should be

2 1 1 6 1 0 # Gauss type functions l=1 set 1 (3p orbital)

Please also note that this is a not very well tested option.
I'm not aware of any successful application using ECPs with CP2K.

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie E-mail: hut...@chem.uzh.ch
Universität Zürich
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: "cp2k" <cp...@googlegroups.com>
From: "Eugene"
Sent by: cp...@googlegroups.com
Date: 02/07/2022 05:30PM
Subject: [CP2K:16557] How to use ECP potential in FORCE_EVAL

--
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/a65b2506-cc62-475b-8514-3ad646aea6a3n%40googlegroups.com.

Eugene

unread,

Feb 7, 2022, 2:02:27 PM2/7/22

to cp2k

Hi Jeurg,

Cool it works. But number of MO is not the same for ECP and GTH. Now for ECP we have:

Total number of - Atomic kinds: 2
- Atoms: 84
- Shell sets: 336

- Shells: 336

Number of Shell cets is not equivalent to number of atoms, while in GTH this is the case. Maybe it is common for ECP. So, it means CP2K Documentation is not complete. Should I give a BUG report?

Warm wishes,
Eugene

понедельник, 7 февраля 2022 г. в 21:54:50 UTC+3, jgh:

hut...@chem.uzh.ch

unread,

Feb 8, 2022, 4:32:42 AM2/8/22

to cp...@googlegroups.com

Hi

basis sets and pseudopotentials are not coupled in the CP2K input.
You can mix (even if it doesn't make sense) whatever you want.

Shell sets (per atom) is the first lone number in the basis set definition.
In your case this is 4, so you get 84*4 Shell sets in total. As, again in
your case, each set has exactly 1 shell, you get the same number of
total shells.
Other basis sets have less sets (e.g. MOLOPT have 1 set), but more shells per
set. Everything is as it should be, there is no bug here.

regards

Juerg Hutter
--------------------------------------------------------------
Juerg Hutter Phone : ++41 44 635 4491
Institut für Chemie E-mail: hut...@chem.uzh.ch
Universität Zürich
Winterthurerstrasse 190
CH-8057 Zürich, Switzerland
---------------------------------------------------------------

-----cp...@googlegroups.com wrote: -----
To: "cp2k" <cp...@googlegroups.com>
From: "Eugene"
Sent by: cp...@googlegroups.com

Date: 02/07/2022 08:02PM
Subject: Re: [CP2K:16561] How to use ECP potential in FORCE_EVAL

To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/32e8d80d-d886-4a51-b040-de7a787533a3n%40googlegroups.com.

Eugene

unread,

Feb 8, 2022, 5:20:54 AM2/8/22

to cp2k

Dear Juerg,

I'm sorry for confusion. I was asking about documentation "bug". From explanation given here: https://www.cp2k.org/basis_sets It is more or less clear that for d-orbitals one have to use 5 numbers in line 9. But nothing said about the case when s- and p-orbitals are given separately.

Anyway I'm really appracieted for you gentle help.

Sincerely,
Eugene

вторник, 8 февраля 2022 г. в 12:32:42 UTC+3, jgh:

Reply all

Reply to author

Forward