[molpro-user] Analytical gradients for DF-MULTI (DF-CASSCF) again

Werner Gyorffy gyorffy at theochem.uni-stuttgart.de
Wed Oct 17 13:10:25 BST 2012


Dear Toshimasa,

The error is due to that the number of closed and core orbitals are the
same ("closed,69" and "core,69") in the df-rs2 calculation which is not
allowed in the current implementation of  DF-CASSCF and DF-CASPT2
gradients. You need to define at least 1 orbital which is closed but not
treated as frozen core: for example, "closed,69" and "core,68". This is
only a technical limitation which will be removed very soon.

Regards,

Werner.

On 10/17/2012 05:16 AM, Toshimasa Ishida wrote:
> Dear all,
> 
> I have trouble with analytical gradients for DF-MULTI (DF-CASSCF) again
> although Dr. Werner Gyorffy showed the nice solution for the H2O system before.
> (Thank you very much, Werner!)
> 
> This time, I used a larger system, and I got a normal result with a conventional basis set.
> On the other hand, when I tried to calculate the same with density-fitting, the final lines were
> -----------------------------------------------------------------------------------
> 1PROGRAM * ALASKA (Gradient of the energy)     Author: R. Lindh
> 
> 
>   RSPT2 Gradient info restored from record    5200.1
> 
>   Number of core orbitals:          69 (  69 )
>   Number of active  orbitals:        2 (   2 )
>   Number of external orbitals:     247 ( 247 )
> 
>   Records on file 6
> 
>   IREC   NAME  TYPE        OFFSET    LENGTH   IMPLEMENTATION   EXT   PREV   PARENT  MPP_STATE
>     1    6000               4096.  5048776.         df          0      0      0      0
> 
>   READM:  RECORD=    6000.6  EXTENSION= 0  OFFSET=    4996280.  ADDRESS=    5000376.  LEN=       52489  IMPLEMENTATION=df
>   CURRENT WRITTEN FILE LENGTH:  5000369.TRYING TO READ UP TO ADDRESS:  5052865.
>   ? Error
>   ? Trying to read beyond end of file
>   ? The problem occurs in readm
> -----------------------------------------------------------------------------------
>   although the rs2 energy (= MCSCF energy ) was normally obtained.
> (The input file and the difference from the non-DF input are shown below)
> 
> Thank you very much for your comments/suggestions in advance.
> 
> Best regards,
> 
> Toshimasa Ishida
> 
> *******************
> The input is
> -----------------------------------------------------------------------------------
>   memory,1600,m
> 
>    file,2,2x2.wfu,new;
>    punch,2x2.pun
>    status,2,clear;
> 
>    angstrom
>    geomtyp=xyz
>    geometry={
>    45
>                                                      Data generated by TSH program
>   C                 0.0347939857        0.0219449914        0.0552879737
>   C                 0.0682919693       -0.0223429907        1.5880452542
>   C                 1.5218992828        0.0066869981        2.0798780216
>   C                 2.3679938845        1.1141854756        1.4368183265
>   C                 1.9890480652        1.6991762025        0.2503968830
>   C                 0.7049996672        1.3066593850       -0.4566787826
>   C                 3.5825013126        1.3999113423        2.1717609768
>   C                 4.3938419343        2.5120098178        2.1789359766
>   C                 5.5782553733        2.6471507543        2.9793885964
>   C                 6.2601540567        3.8711361764        2.9224106271
>   C                 7.4350935023        4.1853950288        3.6303322934
>   C                 8.2085781350        5.3521994839        3.6764672724
>   C                 8.0633572074        6.6018408939        3.0326705713
>   C                 9.0485687435        7.5822894348        3.3107134441
>   C                 9.0966977197        8.8636208317        2.7905326860
>   C                 2.7867326871        2.7616196983       -0.4691367787
>   C                 6.0347161614        1.4875112976        3.8334611976
>   C                 6.9162917476        6.8846857623        2.0934580158
>   N                10.0535402674        9.7641814047        3.0812955504
>   H                 1.9964620625       -0.9668245456        1.8772271177
>   H                 1.5388192782        0.1316739381        3.1701135071
>   H                 8.3415580767        9.2199906637        2.0990700120
>   H                 9.8412473669        7.2956465678        3.9989321202
>   H                 9.0767497313        5.2809655127        4.3273209617
>   H                 7.8315123175        3.3941744045        4.2611209930
>   H                 5.8227692580        4.6171798277        2.2714409317
>   H                 4.1272860601        3.3779154094        1.5854922556
>   H                 3.8520421896        0.6041977142        2.8660006500
>   H                 0.9098875717        1.2246104224       -1.5330012787
>   H                 0.0044389960        2.1553569872       -0.3633648270
>   H                 0.5637217328       -0.8538975970       -0.3452758373
>   H                -0.9936785321       -0.0317199859       -0.3185248498
>   H                -0.4317357972       -0.9220265640        1.9638510748
>   H                -0.4786837773        0.8398586053        1.9950800639
>   H                 2.5068438226        2.8010346798       -1.5257982801
>   H                 2.5920947824        3.7609292293       -0.0526179741
>   H                 3.8653931824        2.5847587858       -0.4107188076
>   H                 6.2177380754        0.6006737182        3.2164574847
>   H                 6.9485857319        1.6945952019        4.3882379340
>   H                 5.2631835251        1.2197034276        4.5651488538
>   H                 6.9353947357        7.8932892872        1.6807502118
>   H                 5.9570961986        6.7612158192        2.6089377719
>   H                 6.9292257384        6.1865500869        1.2486664119
>   H                10.0470732746       10.6881539695        2.6773167379
>   H                10.8111719108        9.5466825086        3.7139642519
>   }
> 
> basis={
>     default,6-31G(d)
>    set,fit              ! defines a basis set called fit
>     default,vdz/jkfit
> }
> {cfit,basis=fit}      ! defines the DF basis set for all calculations
> 
> {df-hf;
>   wf,sym=1,charge=1,spin=0}
> 
>   {df-multi;maxiter,40;
>    occ,71;closed,69;
>    thresh,pspace=10.0;
>    wf,sym=1,charge=1;state,2;
>    }
> 
> {df-rs2;             ! invokes the rs2 programs
>    occ,71;closed,69;core,69;
> noexc;               ! considers only the reference space
> state,1,2}           ! for the ground state
> 
> force
> ---
> 
> The difference from the non-DF input, which lead to a normal result, is
> (diff output)
> ----------------------------------------------------------------------
> 59c59,64
> <   basis=6-31G
> ---
>> basis={
>>     default,6-31G(d)
>>    set,fit              ! defines a basis set called fit
>>     default,vdz/jkfit
>> }
>> {cfit,basis=fit}      ! defines the DF basis set for all calculations
> 61c66
> < {hf;
> ---
>> {df-hf;
> 64c69
> <  {multi;maxiter,40;
> ---
>>   {df-multi;maxiter,40;
> 70c75
> < {rs2;             ! invokes the rs2 programs
> ---
>> {df-rs2;             ! invokes the rs2 programs
> --------------------------------------------------------------------------------------
> 
> ----------------------------------------
> Toshimasa ISHIDA
> ishida at fukui.kyoto-u.ac.jp
> Program-Specific Associate Professor
> Fukui Institute for Fundamental Chemistry, Kyoto Univ.
> 34-4, Takano-nishihirakicho, Sakyo-ku, Kyoto 606-8103, Japan Phone, Fax:
> +81-75-711-7838
> TEL,FAX: 075-711-7838
> ----- Original Message -----
>> From: Werner Gyorffy <gyorffy at theochem.uni-stuttgart.de>
>> To: molpro-user at molpro.net
>> Date: 2012-10-04 22:42:54
>> Subject: Re: [molpro-user] Analytical gradients for DF-MULTI (DF-CASSCF)
>>
>> Dear Toshimasa,
>>
>> State-averaged MCSCF/CASSCF and CASPT2 analytical gradients using the DF
>> approximation are available through the RS2 gradient program. For
>> MCSCF/CASSCF gradient, one must use the NOEXC directive in the rs2 card.
>> For input examples, see the df_casscf_opt and df_caspt2_opt tests in the
>> Molpro/testjobs directory. Your input then would be something like this:
>>
>> ::::::::::::::
>> memory,2,m
>>
>> punch,h2o.pun
>>
>> gthresh,energy=1.d-7
>> geometry={
>> O;               !define z-matrix
>> H1,O,r;
>> H2,O,r,H1,a
>> }
>>
>> r =1.0 ang
>> a = 100.0
>>
>> basis={
>>     default,6-31G(d)
>>    set,fit              ! defines a basis set called fit
>>     default,vqz/jkfit
>> }
>> {cfit,basis=fit}      ! defines the DF basis set for all calculations
>>
>> df-hf;                                      !do scf
>>
>> {df-multi;
>>    maxiter,40;
>>    occ,6;closed,4;
>>    thresh,pspace=10.0;
>>    wf,sym=1,charge=0;state,2;}
>>
>> {df-rs2;             ! invokes the rs2 programs
>> noexc;               ! considers only the reference space
>> state,1,1}           ! for the ground state
>>
>> optg
>> ::::::::::::::
>>
>> These programs are still under development. More flexible input options
>> and a full documentation is coming soon.
>>
>> Regards,
>>
>> Werner.
>>
>> -- 
>>      Dr. Werner Gyrffy
>>
>>      Institut fr Theoretische Chemie
>>      Universitt Stuttgart
>>      Pfaffenwaldring 55
>>      D-70569 Stuttgart
>>      Germany
>>
>>      Email: gyorffy at theochem.uni-stuttgart.de
>>      Phone: +49 711 68569084
>>      Fax: +49 711 68564442
>>      WWW: http://www.theochem.uni-stuttgart.de/
>>
>> On 10/04/2012 04:09 AM, Toshimasa Ishida wrote:
>>> Dear all:
>>>
>>> I tried calculating analytical gradients of DF-MULTI
>>> because Recent Change B.0.12 in the manual describes
>>> that the gradients are now available.
>>>
>>> Unfortunately, the attempt failed, and the last five lines of the output are as follows:
>>>
>>> --------------------------------------------------------------------
>>> Geometry optimization using default procedure for command DF-MULTI
>>>
>>>    Geometry written to block  1 of record 700
>>>    ? Error
>>>    ? Record not found
>>>    ? The problem occurs in readm
>>> --------------------------------------------------------------------
>>>
>>> On the other hand, the following attempts were found to be successful:
>>>
>>> 1. DF-HF optimization with analytical gradients (just 'optg')
>>> 2. DF-MULTI single point calculation (without 'optg')
>>> 3. DF-MULTI optimization with numerical gradients ('optg,numerical')
>>>
>>> It seems the analytical gradients for DF-MULTI are not available yet.
>>> Is that correct?
>>>
>>> The following is the input used for the failed attempt:
>>>
>>> --------------------------------------------------------------------
>>> memory,800,m
>>>
>>> punch,h2o.pun
>>>
>>> gthresh,energy=1.d-7
>>> geometry={
>>> O;               !define z-matrix
>>> H1,O,r;
>>> H2,O,r,H1,a
>>> }
>>>
>>> r =1.0 ang
>>> a = 100.0
>>>
>>> basis=6-31G(d)
>>>
>>> df-hf,df_basis=vqz;                         !do scf
>>>
>>> {df-multi,df_basis=vqz;
>>>    maxiter,40;
>>>    occ,6;closed,4;
>>>    thresh,pspace=10.0;
>>>    wf,sym=1,charge=0;state,2;
>>>    cpmcscf,grad,state=1.1
>>>    }
>>>
>>> optg
>>> --------------------------------------------------------------------
>>>
>>> Thank you very much for your comments and suggestions in advance.
>>>
>>> Best regards,
>>>
>>> Toshimasa Ishida
>>> -------------------------------------------
>>> Toshimasa ISHIDA
>>> ishida at fukui.kyoto-u.ac.jp
>>> Program-Specific Associate Professor
>>> Fukui Institute for Fundamental Chemistry, Kyoto Univ.
>>> 34-4, Takano-nishihirakicho, Sakyo-ku, Kyoto 606-8103, Japan Phone, Fax:
>>> +81-75-711-7838
>>> _______________________________________________
>>> Molpro-user mailing list
>>> Molpro-user at molpro.net
>>> http://www.molpro.net/mailman/listinfo/molpro-user
>>>
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net
>> http://www.molpro.net/mailman/listinfo/molpro-user
> 
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net
> http://www.molpro.net/mailman/listinfo/molpro-user
> 



More information about the Molpro-user mailing list