[molpro-user] DF-LCCSD(T) with core electrons

Lee E.P.F. E.P.Lee at soton.ac.uk
Tue Sep 6 23:14:48 CEST 2016

Dear MOLPRO users,

We want to use the DF-LCCSD(T) method to do some calns on some large alkali metal complexes. For example, with Li complexes, Li is usually Li+ in the complex. In this connection, the formally Li 1s2 core electrons are now become valence. We therefore have employed the cc-pwCVDZ basis set for Li and try to include Li 1s2 in the correlation by using the core key word. In our case, the default core is 22. We have checked that orbital 22.1 is Li 1s2 from the DF-HF output. So, we use core 21 with DF-LCCSD(T). However in the output, we have

Deleting projected core orbital  826.1 (LI 1s)   Norm=  0.27D-02

and also

 Ranking of atoms according to Loewdin charges

 Orbital domains

   Orb.   Atom     Charge      Crit.
  22.1    9 C       1.02       0.000
         12 H       0.67       0.989
  23.1    9 C       1.02       0.000
         10 H       0.66       0.988
  24.1    9 C       1.01       0.000
         11 H       0.66       0.988

In the orbital domain, the only Li orbital included is

  53.1   60 LI      1.88       0.989

Have the Li 1s2 electrons been properly correlated ? Can a local correlation method deal with core electrons generally and/or in the case of Li+ specifically ? In the case of Li+, the Li 1s2 e's may be considered as quite deep down, but for heavier alkali metal, it has been shown that, e.g. for Na+, 2s2 2p6 are important. Would using just the core keyword deal with all situations ? Many thanks for your advice in advance.

Best Regards,
Edmond Lee

From: Dyke J.M.
Sent: 23 August 2016 11:42
To: Lee E.P.F.
Subject: FW: M(Me4cyclen)Solv+ project (Single point calculations)

From: Prof Ponnadurai Ramasami [mailto:ramchemi at intnet.mu]
Sent: 19 August 2016 12:14
To: Dyke J.M.; jmdyke99 at gmail.com
Cc: 'Lydia Rhyman'; hbhakhoa at gmail.com
Subject: FW: M(Me4cyclen)Solv+ project (Single point calculations)

Dear John,

Please find attached files relevant to the single point computations.

Kind regards,

More information about the Molpro-user mailing list