[molpro-user] Analytic geometry optimization with density-fitting CASSCF by RS2 did not work

[Werner Győrffy]

Dear MJ Wu,

You have performed a transition state search (optg,root=2) in your 
calculation, which requires calculating the Hessian. The Hessian can be 
obtained by computing second derivatives of the energy with respect to 
nuclear displacements. The gradients, which are first derivatives of the 
energy, were calculated by using analytical gradients. The Hessian 
however was calculated by using numerical first derivatives of the 
analytical gradients. Currently, this is the only way in Molpro to 
compute the DF-CASSCF Hessian.

Regards,

Werner.

On 09/19/2016 07:30 PM, Kenny W wrote:
> Dear Molpro Experts,
>
> I tried to do geometry optimization at df-CASSCF level, and wished the
> job to run using analytic gradient via df-RS2 with "noexc"
> keyword:(reference: http://www.molpro.net/pipermail/molpro-user/2012-October/005092.html
> )
> However, the geometry optimization still proceeded with numerical
> gradient. Here is my input:
>
> ***,df-CAS(12,12) optimization
> memory,15000,m
>
> cartesian
> basis=6-31G*
> geomtyp=xyz
> geometry={
> 39
> TS optimization
>  C                 -4.85773400    0.51429900    0.42945500
>  .... atomic coordinates follows ...
>  H                  0.73880900    2.27415000   -0.56530600
> }
> ! orbital from previous df-CAS/6-31G*
> {matrop
> read,oldorbitals,file=radical_df4.orb
> save,oldorbitals,2140.2,orbitals}
>
> {df-multi,DF_BASIS=vdz/JKFIT
> start,2140.2
> close,76
> occ,88
> wf,164,1,0
> natorb,ci,print
> orbital,2141.2
> maxiter,39
> }
>
> ! change basis to VDZ with density fitting
> basis={
>    default,vdz
>   set,fit             ! defines a basis set called fit
>    default,vdz/JKFIT} ! here uses slash in the basis name
> {cfit,basis=fit}      ! defines the DF basis set
> spherical
>
> {df-multi
> start,2141.2
> close,76
> occ,88
> wf,164,1,0
> natorb,ci,print
> orbital,2142.2
> maxiter,39
> }
>
> ! df-CAS analytic gradient using df-RS2 with keyword noexc
> {df-RS2,thrden=1.0d-11,thrvar=1.0d-11,maxit=100,maxiti=100
> core,75              ! this should be redundant with noexc
> occ,88
> close,76
> noexc;               ! this is just for analytic grad
>  wf,164,1,0;
> state,1,1
> orbit,2142.2
> }
>
> optg,root=2          ! optimize for transition state
>
> Everything went fine until df-RS2 finished, and the RSPT2 energy indeed
> equals to the CASSCF reference energy. But while starting geometry
> optimization, I found in the output file that the calculation still used
> numeric gradient.
>
> May I know what went wrong with the input file? The tail of the output
> file follows. Thank you very much for your help!
>
> Sincerely,
> MJ Wu
>
>  *Reference energy                   -1293.32140467*
>  Nuclear energy                      1881.75751251
>  Kinetic energy                      1292.96788162
>  One electron energy                -5511.10187574
>  Two electron energy                 2336.02295856
>  Virial quotient                       -1.00027342
>  Correlation energy                     0.00000000
>  *!RSPT2 STATE 1.1 Energy            -1293.32140466*6142
>
>  !RSPT2 STATE 1.1 Dipole moment        -1.75999116    -0.81167791
> 0.50501499
>  Dipole moment /Debye                  -4.47315833    -2.06294434
> 1.28353600
>
>  Hylleraas correlation energy:          0.00000000
>  *Total energy:                      -1293.32140467*
>
>  Solving MCSCF z-vector equations for state  1.1, requested accuracy=
>    0.10D-05
>  Convergence reached in   4 iterations (  4 expansion vectors),
> accuracy=  0.65D-06
>
>  Reference dipole moments:             -1.75999116    -0.81167791
> 0.50501499
>  !RSPT2 dipole moments (ED):           -1.76000654    -0.81168150
> 0.50501620
>
>  RSPT2 Gradient info saved on record         5200.1
>
>  CPU-time for rhs:          703.84 sec
>  CPU-time for cpmcscf:     3765.26 sec
>  CPU-time for densities:      2.35 sec
>
>
>  **********************************************************************************************************************************
>  DATASETS  * FILE   NREC   LENGTH (MB)   RECORD NAMES
>               1      25     7557.78       500      610      700      900
>      950      970     1001      129      960     1100
>                                           VAR    BASINP    GEOM
>  SYMINP    ZMAT    AOBASIS   BASIS     P2S    ABASIS      S
>                                          1400     1410     1200     1210
>     1080     1600     1650     1700     1380     5200
>                                            T        V       H0       H01
>     AOSYM     SMH    MOLCAS    OPER     JKOP    RS2GRAD
>                                          5210     7009    90002    90003
>     7005
>                                         RS2GRAD             ITF      ITF
>
>               2       7       18.73       700     1000      520     2140
>     2141     1001     2142
>                                          GEOM     BASIS   MCVARS
> MATROP    MCSCF    BASIS    MCSCF
>
>  PROGRAMS   *        TOTAL       RS2     MULTI       INT     MULTI
>  MATROP       INT
>  CPU TIMES  *     42786.30   5427.08  21523.68     13.74  15548.16
>  2.63    266.23
>  REAL TIME  *      4057.84 SEC
>  DISK USED  *        78.76 GB
>  **********************************************************************************************************************************
>
>
>  PROGRAM * OPT (Geometry optimization)     Authors: F. Eckert and H.-J.
> Werner
>
>
>  Geometry optimization using default procedure for command DF-RS2
>
>  Geometry written to block  1 of record 700
>
> * Generated 111 symmetrical displacements*
>
> * Numerically approximating hessian using forward gradient differences*
>
>  Task list generated. Total number of displacements:    111
>
>    11 tasks completed, CPU=3d 16h 28m 23s Elapsed=9h 25m 6s
>    22 tasks completed, CPU=7d 2h 25m 20s Elapsed=17h 49m 50s
>
>
>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
>