UCCSD(T) issue

[anthony.scott at anu.edu.au]

Fellow Molpro users,

A researcher here has found some interesting issues with Molpro2002.6.

She starts with a single geometry, open shell.  On occasion she finds 
that the HF part converges to a higher energy relative to closely 
related geometries.  She seems to be able to overcome this by employing 
a dft guess for the uccsd(t) calculation.

Now,

(1) If she does: dft, hf, uccsd(t) calculation she obtains one answer.

(2) If she does: uhf, hf, uccsd(t) she gets an answer that sometimes is 
not the same as in (1) usually for higher energy type structures.

(3) If she does: hf, hf, uccsd(t) she get an answer that is different 
again from (1) on occasion,  by up to 0.01 Hartree.

Could anyone help out and tell us what is going on here.  SCF 
convergence is set to 10^-12.

Thanks heaps.

tony
_______________________________________________________
Dr. Anthony P. Scott,
Computational Quantum Chemistry Gp,	Office Ph.: 	61-2-6125-3573
Research School of Chemistry,		Dept. Ph.:  	61-2-6125-3637
Australian National University,			Fax: 			61-2-6125-0750
Canberra, ACT 0200,				
AUSTRALIA.

E-mail: 	Anthony.Scott at anu.edu.au
_______________________________________________________