[molpro-user] UCCSD(T) numerical gradient failing in rhfpro

[Matt Addicoat]

Dear Molpro-users,

I have used the following procedure to successfully optimise the 2a'
state of AuC2:

foo={
{rhf,maxdis=10,iptyp='KAIN',maxit=500,nitord=5; shift,-1.0,-0.5}
uccsd(t),maxit=60}                        
optg,proc=foo;

The reason I'm getting a bit cavalier with my scf options is that one of
the energy calculations keeps failing (no convergence) in rhfpro. I
can't see any obvious oscillatory behaviour and neighbouring points
converge to what I expect.

I'm now trying to do a relaxed scan of the Au-C-C angle and for some
points, even the above options fail. I tried the following procedure,
which works as a single point calculation, but it fails at the same
gradient task and inspection of the logfile tells me that my
save/orbital directives aren't being respected.

foo={
{rks,b-lyp;save,2100.2}
{rhf;nitord=1; orbital,2100.2}
uccsd(t),maxit=40}
optg,proc=foo;

Can anyone advise me on a more robust way to complete this optimisation?

Cheers,
Matt