(fwd) adding numerical gradients

[H. -J. Werner]

Hi,
This case is not implemented. When analytical gradients are
used the ADD command is processed by the gradient program,
but in this case you never get there. It would certainly be
possible to hack something into opt/numforces.f, but I have
no time to do this soon. You can look in forces/kraft.f
to see what has to be done (seach for addfac). Should you get
it working please send the modifications so that we can
include them into the master version
Best regards
Joachim Werner




----- Forwarded message from Jochen Küpper <jochen at unc.edu> -----

From: "Jochen Küpper" <jochen at unc.edu>
To: molpro-user <molpro-user at stchem.bham.ac.uk>
Subject: adding numerical gradients
Date: 09 Jul 2002 15:12:58 -0400
Message-ID: <ly8z4k1zr9.fsf at bock.chem.unc.edu>
User-Agent: XEmacs/Gnus
Organization: University of North Carolina

Dear All,

we are having problems doing counterpoise-corrected geometry
optimizations of open-shell systems with molpro-2002.3.

In section 35.2.22 of the manual this is described for closed shell
systems and we have been able to perform such calculations for many
closed shell systems.

When one wants to do CP corrected geometry otpimizations for
open-shell systems (i.e. RCCSD(T)) numercial gradients have to be
used.  This leads to the error message
,----
| ILLEGAL COMMAND  ADD 
`----
after the calculation of the gradients.

The error can be reproduced by substituting the "forces" commands in
the example from the manual by
,----
| forces,numerical,startcmd=hf
`----
The complete input file is attached.

Is there any way to perform counterpoise-corrected geometry
optimizations on open-shell systems (RCCSD(T)) where it is necessary
to use numerical gradients?

Greetings,
Jochen
-- 
University of North Carolina                       phone: +1-919-962-4403
Department of Chemistry                            phone: +1-919-962-1579
Venable Hall CB#3290 (Kenan C148)                    fax: +1-919-843-6041
Chapel Hill, NC 27599, USA                            GnuPG key: 44BCCD8E

***,HF dimer mp2/CP optimization

basis=avdz

maxit=20                !max number of iterations

text,OPTIMIZED VALUES OF GEOMETRY VARIABLES 

RFF=      5.31431160
R1=       1.75768738 
R2 =      1.75298524  
THETA1 =  7.03780227 
THETA2 = 111.25930975

geometry={x;noorient    !noorient must be specified since gradients are added
      f1
      f2  f1  rff
      h1  f1  r1   f2  theta1
      h2  f2  r2   f1  theta2  h1  180.}

do iter=1,maxit         !optimization loop

text, CALCULATION AT LARGE SEPARATION

rff_save=rff            !save current rff distance
rff=1000                !calculation at large separation

text, HF1 MONOMER

dummy,f2,h2;            !second hf is now dummy
hf;                     !scf for first monomer
mp2;                    !mp2 for first monomer
ehf1inf=energy          !save mp2 energy in variable
forces,numerical,startcmd=hf

text, HF2 MONOMER

dummy,f1,h1;            !first hf is now dummy
hf;                     !scf for second monomer
mp2;                    !mp2 for second monomer
ehf2inf=energy          !save mp2 energy in variable
forces,numerical,startcmd=hf
add,1                   !add from previous gradient

rff=rff_save            !reset HF - HF distance to current value

text, HF1 CP MONOMER

dummy,f2,h2;            !second hf is now dummy
hf;                     !scf for first monomer
mp2;                    !mp2 for first monomer
ehf1=energy             !save mp2 energy in variable
forces,numerical,startcmd=hf
add,-1                  !subtract from previous gradient

text, HF2 CP MONOMER

dummy,f1,h1;            !first hf is now dummy
hf;                     !scf for second monomer
mp2;                    !mp2 for second monomer
ehf2=energy             !save mp2 energy in variable
forces,numerical,startcmd=hf
add,-1                  !subtract from previous gradient

dummy                   !reset dummies

text, DIMER CALCULATION
hf;                     !scf for dimer
mp2;                    !mp2 for dimer
edimer=energy           !save mp2 energy in variable
forces,numerical,startcmd=hf
add,1                   !add to previous gradient

decpc1=ehf1inf-ehf1     !counterpoise correction for first monomer
decpc2=ehf2inf-ehf2     !counterpoise correction for second monomer
decpc=decpc1+decpc2     !total counterpoise correction
de=edimer-ehf1-ehf2     !Interaction energy relative to unrelaxed monomers
etot=edimer+decpc       !total cpc corrected energy

opt;                    !find next geometry
show,optconv            !show convergence parameter
if(optconv.lt.0.0003) goto,lab:  !test for convergence



enddo            !end of optimzation loop

lab:

text, compute optimized monomer energy 

rhf=r1
geometry={h1
          F1,H1,rhf}
hf;                     !scf for relaxed monomer
mp2;                    !mp2 for relaxed monomer
ehf=energy              !save mp2 energy in variable
optg                    !optimize monomer structure

text, compute interaction energy including relaxation

erelax = 2*ehf - ehf1inf - ehf2inf   !relaxation energy
derel=de-erelax         !interaction energy relative to relaxed monomoers

text, results 

show,erelax,de,derel,r1,r2,rhf,rff,theta1,theta2

decpc=decpc*tocm        !total counterpoise correction in cm-1
erelax=erelax*tocm      !relaxation energy in cm-1
de=de*tocm              !interaction energy relative to unrelaxed monomers in cm-1
derel=derel*tocm        !interaction energy relative to relaxed monomers in cm'1


---;




----- End forwarded message -----

-- 
Prof. Hans-Joachim Werner
Institute for Theoretical Chemistry
University of Stuttgart
Pfaffenwaldring 55
D-70569 Stuttgart, Germany
Tel.: (0049) 711 / 685 4400
Fax.: (0049) 711 / 685 4442
e-mail: werner at theochem.uni-stuttgart.de