[molpro-user] cp-corr. Potential DF-LCCSD(T0), numerical accuracy

[Berger Raphael]

Dear Readers,

when trying to calculate a cp-corrected potential curve
in a close region around a uncorrected minimum
it seems to give a "noisy" curve for the dE_cp using
only the default settings. The cp-correction at the
equlibrium geometry is about 8*10^-4 au.
The level is DF-LCCSD(T0)/SARC-TZVPP with DKH3
integrals.

I saw in the manual that the BSSE examples use 
{hf;accu,16}. My question is, what other non-default settings
could be important to get a smooth cp corrected potential
curve for this method?
My input is basically:

geometry={Au1;H2,Au1,rAuH;Au3,H2,rAuH,Au1,wAuHAu(i);
}
rAuH=1.6900 angstrom
wAuHAu=[100.7,100.9,100.4,100.6,100.8,101.0]
set,charge=1
basis={
...
}
gdirect
do i=1,#wAuHAu
dkroll=1
dkho=3
hf
{df-lccsd(t0),df_basis=orcadf;}
e_tot(i)=energy
text cp calc AuH fragment
dummy,Au3
set,charge=0
dkroll=1
dkho=3

hf
{df-lccsd(t0),df_basis=orcadf;}
e_AuH(i)=energy
text cp calc Au+ fragment
dummy,Au1,H2
set,charge=1
dkroll=1
dkho=3
hf
{df-lccsd(t0),df_basis=orcadf;}
e_Au_plus(i)=energy

e_corr(i)=e_tot(i)-e_AuH(i)-e_Au_plus(i)
enddo



(I use molpro 2008.1_mpp Linux x86_64)


Thank You for reading
 	Raphael Berger