Generating start guess for embedded cluster

[tonner]

Hi,

I am currently testing the interesting PEECM feature in 6.0.2 and my main problem in the moment is generating a start guess.
The system is a Ti36O72-cluster embedded in Ti-ECPs.

The problem starts if I assign no basis sets to my Ti-cores which are just represented by ECPs (and the O-atoms represented by point charges) and try to generate an EHT-guess:
> OCCUP. NUMBER ASSIGNMENT CAN NOT BE CARRIED OUT
 SINCE THERE ARE LESS ORBITALS THAN ELECTRONS
 ALL FILES WILL BE REMOVED AND INPUT TERMINATED
 SORRY !

This is reproducible with the test-file TURBOTEST/dscf/long/Ca4F19.PC.HF if one removes the mos contained in the control file and tries to generate an EHT-guess.

The only way I found up to now is starting without a guess (scfmo = none) but then it takes about 30-40 iterations to reach a reasonable wavefunction.

Is there any more clever way of generating a guess for a cluster embedded in ECPs?

Thanks and regards
Ralf.

[flabia]

Hi,

I know this comes as a late reply, but just some days ago I got stuck in the same problem, and as I guess this can be a little tricky also for other users, I suggest you to solve it with this procedure:

in define, when you call eht, you are asked to enter the molecular charge: as you have some cations which are represented by ECP, you have to give a charge equal to the number of ECPs times the electrons for each atom represented by ECPs.

For example, if you have 76 Zn atoms represented by ECPs with 10 core electrons (e.g. the library ECP ecp-10-mdf), you'll have to enter a charge of
76*(Z-10), where Z is the atomic number of Zn (30 electrons), i.e. you have a charge of 1520.

NB: in your coord file, you only have to put the QM part and the ECPs, DO NOT put the point charges of the anions in the isolating shell!!

hope this helps

Claudia