[molpro-user] Multi with F12?

Kirk Peterson kipeters at wsu.edu
Wed Sep 29 19:35:21 BST 2010


Hi Terry,

here's a workaround below.  The problem is that optg has some defaults/protections built-in that attempts
to minimize problems that might arise from using a guess procedure as you want to use.  The way around
it is to use the minimize directive to optimize your geometry since it is a barebones optimizer.  In my example
below I just take the defaults for convergence on the gradient, so you might want to play with that.

regards,

Kirk


basis=VDZ-F12
symmetry NoSym
Angstrom
geometry={
 C
 N C rCN
 H C rCH N aHCN
 O C R N aOCN H dNCOH }

R=1.6
rCN=1.23
rCH=1.08
aHCN=132.0
aOCN=114.5
dNCOH=179.98

int

{minimize,eval,rCN,rCH,aHCN,aOCN,dNCOH
method,bfgs,proc=Multi-RHF-CCSD}

proc multi-rhf-ccsd

  {multi; occ 12; closed 10; WF 22 1 2; accuracy step=0.01; maxiter 40; canonical,3100.2}
  {RHF;start,3100.2}
  {RCCSD-F12A,ri_basis=vdz-f12/optri}
  eval=energy

endproc
---

On Sep 28, 2010, at 8:59 PM, Terry Frankcombe wrote:

> Hi Kirk, molpro people
> 
> I've played around with a number of things, but am yet to get this
> working.
> 
> The closest I've gotten so far is:
> 
> Multi-RHF-CCSD={
>  {multi; occ 12; closed 10; WF 22 1 2; accuracy step=0.01; maxiter 40}
>  {RHF}
>  {RCCSD-F12A,ri_basis=vdz-f12/optri} }
> 
> Multi-RHF-CCSD
> {OptG,numerical,proc=Multi-RHF-CCSD; inactive R}
> 
> 
> The problem is where the orbitals come from and go to.
> 
> As I've written it here, the first multi dumps to 2140.2, that's read by
> RHF, which dumps to 2100.2, which is read by RCCSD.
> 
> When OptG starts multi and RHF read the previous orbitals on 2140.2 and
> 2100.2, respectively, so I've gained nothing (and the RHF fails during
> the stepping for the gradient).
> 
> If I specify "orbital 2222.2" for multi and RHF then again it works for
> the initial energy, but once OptG gets going the orbitals from multi are
> dumped in 2222.1, then the subsequent RHF reads from 2222.2.  Again,
> I've gained nothing.  If I try to dump in 2222.1 OptG tells me off and
> stops.
> 
> If I use "start 2222.1" for the RHF it gets ignored and orbitals are
> read from 2222.2 in the OptG.
> 
> All this is with 2009.1.  I've attached a copy of the input.  (Yes, in
> this system there is a low-lying excited state, which is likely
> contributing to my convergence problems.)
> 
> If there are any suggestions about how to get this to work, I'll be
> happy to hear them.
> 
> Regards
> Terry
> 
> 
> 
> On Thu, 2010-09-23 at 08:25 -0700, Kirk Peterson wrote:
>> Hi Terry,
>> 
>> Sorry about leading you astray, I thought this now worked but I just tried with 2010.2 and it doesn't find the
>> Fock matrix data.  The only workaround I can think might work is to use a procedure to carry out your
>> energy evaluation and in there directly follow your multi calculation with a regular hf, making sure the latter
>> reads the orbitals from multi.
>> 
>> -Kirk
>> 
>> On Sep 23, 2010, at 12:19 AM, Terry Frankcombe wrote:
>> 
>>> On Wed, 2010-09-22 at 08:02 -0700, Kirk Peterson wrote:
>>>> Hi Terry,
>>>> For using multi for your HF, just invoke multi where you would normally call hf but include
>>>> occ and closed cards to restrict it to a single determinant.  Also output the orbitals as type
>>>> canonical.  In the rccsd-f12a calculation, specify the record of these orbitals to read in and you might have
>>>> to also include a "ignore_error" option on the orbital directive since usually the ccsd(t) code doesn't
>>>> like what looks like non-canonical orbitals for the (t) step.
>>> 
>>> Hi Kirk
>>> 
>>> Thanks for that.  It looks like I'm getting close, but not there yet.
>>> 
>>> With 2009.1:
>>> 
>>> {multi; occ 12; closed 10; WF 22 1 2; canonical 2140.2}
>>> {RCCSD-F12A,ri_basis=vdz-f12/optri; orbital 2140.2,canonical}
>>> 
>>> In the RCCSD I get:
>>> 
>>> Augmented Hessian optimization of geminal fit. Trust ratio= 0.40000
>>> Convergence reached after   2 iterations. Final gradient= 8.43D-16, Step= 4.27D-06, Delta= 1.28D-09
>>> 
>>> Alpha:                 0.19532     0.81920     2.85917     9.50073    35.69989   197.79328
>>> Coeff:                 0.27070     0.30552     0.18297     0.10986     0.06810     0.04224
>>> 
>>> 
>>> 
>>> ?ERROR IN READ_DUMP: OCC SET= 1 NOT FOUND IN RECORD    2140.2
>>> 
>>> THE RECORD CONTAINS THE FOLLOWING MCSCF DATA:
>>> ORBITALS/MULTI              SET= 1  STATE=  1.1  MS2=2  NELEC=  2
>>> DENSITY/CHARGE              SET= 1  STATE=  1.1  MS2=2  NELEC=  2
>>> ORBITALS/CANONICAL          SET= 2  STATE=  1.1  MS2=2  NELEC=  2
>>> GROUP/CANONICAL             SET= 1  STATE=  1.1  MS2=2  NELEC=  2
>>> OCC/CANONICAL               SET= 2  STATE=  1.1  MS2=2  NELEC=  2
>>> EIG/CANONICAL               SET= 2  STATE=  1.1  MS2=2  NELEC=  2
>>> 
>>> GLOBAL ERROR fehler on processor   0
>>> 
>>> 
>>> 
>>> I get the same whether I give the canonical option, the set=2 option, or
>>> no option to the orbital directive.
>>> 
>>> How can I get RCCSD to read some other occupations, or get the
>>> occupations into set 2?
>>> 
>>> Ciao
>>> Terry
>>> 
>>> 
>> 
> -- 
> Dr. Terry Frankcombe
> Research School of Chemistry, Australian National University
> Ph: (+61) 0417 163 509    Skype: terry.frankcombe
> <opt3.inp>




More information about the Molpro-user mailing list