[molpro-user] Numerical Freq calculations using F12-CCSD(T)

Sat Jan 21 08:58:07 CET 2017

Dear John,

When you use CABS formalism, you do not need to add the AO basis to the 
RI basis set in the input. It is enough to define the auxiliary/RI basis 
set, the AO basis is added automatically. The actual RI basis set 
considered in the calculation is constructed as AO + CABS where CABS = 
[auxiliary basis] - AO. The auxiliary basis is the one you define as RI 
set in the input. On the other hand, if you choose not to use the 
complementary auxiliary basis set approach, you should add the AO 
manually if the auxiliary basis set is developed to augment the AO 
basis. You can always use the compact form of explicit basis set 
definition if you are happy with the default contraction scheme. You 
never have to define basis sets with exponents and contraction 
coefficients for built-in basis sets unless you want to modify it. To 
form union of basis sets, by staying with the Dunning AO basis set + the 
corresponding OPTRI basis example,

spd,O,aug-cc-pVDZ;c
spdfg,O,aug-cc-pVDZ/OPTRI;c

When you however choose, for example, the JKFIT basis as an RI basis, 
you do not need to add the AO basis, as the JKFIT basis is large enough 
to describe the AO basis "completely". All depends what kind of basis 
set you use. Just to be sure, I would check the the papers about the 
aug-cc-pwCVDZ RI sets. I have no experience with that.

 > "Encountered below-occupied eigenvalues in virtual block. This may
 > indicate serious deficiencies in the orbital or JKfit-basis sets
 > (basis_exch). To ignore this error, use explicit,ignore_bad_singl=1."
 > However, when I extracted the problematic point (the structure that
 > caused the printed error message) during the frequency calculation,
 > and performed a single point calculation, there was no error message
 > showed up. The calculation finished smoothly.

The error might be related to the CABS construction and freezing 
auxiliary basis sets in numerical frequency calculation. The CABS 
singles correction is a PT2 singles correction to the CA space. The PT2 
is ill-conditioned in the problematic case. To circumvent the problem 
one should avoid using union a basis sets with redundant functions.

Regards,

Werner.

On 01/18/2017 01:13 AM, John Travers wrote:
> Dear Werner
>
> Thank you for the suggestions! I guess it is related to the choice of
> the basis. Since I used a aug-cc-pcwVQZ basis, there is no corresponding
> JK fit ones. I used an aug-cc-pV5Z JKFIT (see attached input). Do you
> have other suggestions for the choice of ri, df, and jk basis for the
> aug-cc-pcwVQZ orbital basis? BTW, the molpro version I used is 2012.1.
> Thanks again!
>
> Best wishes
>
> John
>
>
> On Tuesday, January 17, 2017 8:13 AM, Werner Győrffy
> <gyorffy at theochem.uni-stuttgart.de> wrote:
>
>
> Dear John,
>
> There might be some problem with the choice of the basis sets you use.
> It is recommended to use the optri basis sets, for example, for
> basis=avtz use ccsd(t)-f12b,ri_basis=avtz/optri. Could you please
> provide an input for your calculation failed?
>
> Regards,
>
> Werner.
>
> On 01/16/2017 05:41 PM, John Travers wrote:
>> Hi
>>
>> I was trying to do a numerical frequency calculation using f12-ccsd(t)
>> in c1 symmetry for a small molecule. The program always produces some
>> warning messages like : "Poor overlap with frozen basis set. May have
>> deleted the wrong vectors!
>>  ? The problem occurs in SvdConstructOrthoBasis"
>>
>> When checked the detailed output,  I also found some possible numerical
>> errors, something looks like :
>>
>>  Construction of CABS:
>>  Pseudo-inverse stability              NaN
>>  Smallest eigenvalue of S        -3.71E-13  (threshold= 1.00E-08)
>>  Ratio eigmin/eigmax              -3.71E-13  (threshold= 1.00E-09)
>>  Smallest eigenvalue of S kept    -3.71E-13  (threshold=-3.71E-13, 0
>> functions deleted, 658 kept)
>>  Maximum overlap with frozen deleted vectors: (nDel=  44)
>>    Vector:        1      2      3      4      5      6      7
>>    8      9      10      11      12      13      14      15      16
>>  17      18      51      25      67      44      36      35      43
>>  32      40      38      33      39      55      52      34      22
>>  37      21      42      50      20      41      74      19      53
>>  54      26      23    231      28      24      95      93      27
>> 158      75      61    123      68    108      31      63    163
>> 121    203    152      62      29    157      45      94      72
>> 97      77      65    204    103      30    115    126      66
>> 250      73    131    101    162    116      91    200    122
>>    Overlap:      NaN    NaN    NaN    NaN    NaN    NaN    NaN
>>  NaN    NaN    NaN    NaN    NaN    NaN    NaN    NaN    NaN
>>  NaN    NaN  1.436  1.248  1.102  1.000  0.999  0.997  0.996
>> 0.994  0.985  0.983  0.973  0.964  0.962  0.912  0.896  0.871
>> 0.858  0.853  0.827  0.825  0.742  0.596  0.594  0.537  0.523
>> 0.485  0.465  0.386  0.375  0.373  0.346  0.327  0.321  0.310
>> 0.307  0.268  0.259  0.257  0.249  0.247  0.216  0.204  0.194
>> 0.191  0.183  0.166  0.164  0.159  0.135  0.134  0.130  0.126
>> 0.118  0.114  0.110  0.100  0.095  0.093  0.092  0.092  0.091
>> 0.089  0.080  0.079  0.079  0.078  0.076  0.075  0.072  0.070
>>
>>
>> My questions are related to these "NaN" and warnings. Do these matter
>> and how to avoid this? Thanks!
>>
>> Best wishes
>>
>>
>> John
>
>>
>>
>>
>>
>>
>> _______________________________________________
>> Molpro-user mailing list
>> Molpro-user at molpro.net <mailto:Molpro-user at molpro.net>
>> http://www.molpro.net/mailman/listinfo/molpro-user
>>
> _______________________________________________
> Molpro-user mailing list
> Molpro-user at molpro.net <mailto:Molpro-user at molpro.net>
> http://www.molpro.net/mailman/listinfo/molpro-user
>
>
>