[molpro-user] Numerical Freq calculations using F12-CCSD(T)
Werner Győrffy
gyorffy at theochem.uni-stuttgart.de
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
>
>
>
More information about the Molpro-user
mailing list