Andrei V Scherbinin
andrei.scherb@gmail.com
If there are many atoms with AO basis set on them, some high-L contributions effectively arise from
AOs centered on the neighbouring atoms. As a result, truncated ECP is not identical to the original
one designed by the authors. On the other hand, ECPs are usually derived so as to reproduce atomic
rather than molecular valence states. That is, intrusion of valence AO tails from neighbouring atoms
into the core region is not properly taken care of, whatever ECP is used, either "trancated" or the
"original" one. All we hope is that such tails are negligible...
On Tue Jan 12 '10 2:23am, Richard wrote
---------------------------------------
>Thanks for correcting me on this - I've previously used both types of truncation but never understood which was the right one, since the results from a few tests (energies and structures of some closed-shell 3rd row TM complexes) were almost identical.
>BTW, do you know under what circumstances one really does need the H-term - open shells with high L valence functions perhaps ? (just guessing here..)
>Richard
>
>
>On Mon Jan 11 '10 10:19pm, Andrei V Scherbinin wrote
>----------------------------------------------------
>>I'm afraid such a truncation of G-H-term is not a correct procedure in general,
>>and the results may be consistent only occasionally.
>>A regular way to cut off the H harmonics from the ECP is as follows.
>>The key idea is that one has to omit the (identically zero in the folow-up case!)
>>H-harmonic term. Then the G-H-potential is the new isotropic (i.e. projector-free)
>>G-potential, which should be subtracted from all the semilocal potentials corresponding
>>to L < 4.
>>I give just an example of Stuttgart RSC ECP for Ce atom, before and after the truncation:
>>Before:
>>CE-ECP GEN 28 5
>> 1 ----- H POTENTIAL -----
>> 0.00000000 2 1.00000000
>> 1 ----- S-H POTENTIAL -----
>> 580.08345700 2 20.13782900
>> 1 ----- P-H POTENTIAL -----
>> 310.30283300 2 15.99848200
>> 1 ----- D-H POTENTIAL -----
>> 167.81394400 2 14.97418700
>> 1 ----- F-H POTENTIAL -----
>> -49.39022900 2 23.40245500
>> 1 ----- G-H POTENTIAL -----
>> -21.33187900 2 16.57055300
>>After:
>>
>>
>>CE-ECP GEN 28 4
>> 1 ----- G POTENTIAL -----
>> -21.33187900 2 16.57055300
>> 2 ----- S-G POTENTIAL -----
>> 580.08345700 2 20.13782900
>> 21.33187900 2 16.57055300
>> 2 ----- P-G POTENTIAL -----
>> 310.30283300 2 15.99848200
>> 21.33187900 2 16.57055300
>> 2 ----- D-G POTENTIAL -----
>> 167.81394400 2 14.97418700
>> 21.33187900 2 16.57055300
>> 2 ----- F-G POTENTIAL -----
>> -49.39022900 2 23.40245500
>> 21.33187900 2 16.57055300
>>The latter one may be safely inserted into the $ECP group
>>for both US GAMESS and Firefly.
>>
>>
>>
>>On Sun Jan 10 '10 3:33am, Richard wrote
>>---------------------------------------
>>>I don't know about gamess-us, but one thing you can try is running a similar job in Firefly with and without the G-H term (reducing lmax by 1 in the latter case). If the differences are small enough, then it might be alright to just use the ECP without the G-H term in gamess-us.
>>>Richard
>>>
>>>
>>>On Fri Jan 8 '10 9:02pm, sanya wrote
>>>------------------------------------
>>>>Did I get it right that H-ECPs will work automatically in the next versions of Firefly? Great!
>>>>By the way, what is the situation with H-ECPs in GAMESS-US? I couldn't find anything from the manual...
>>