### Author Topic: RI-approximation in Turbomole  (Read 5372 times)

#### zahra

• Newbie
• Posts: 4
• Karma: +0/-0
##### RI-approximation in Turbomole
« on: February 17, 2017, 03:23:44 PM »
is implementation of RI- approximation in Turbomole software different with implementation of  RI-approximation in Gaussian software with using the same auxiliary basis sets?

#### gunnar.schmitz

• Developers
• Jr. Member
• Posts: 14
• Karma: +0/-0
##### Re: RI-approximation in Turbomole
« Reply #1 on: February 19, 2017, 06:30:44 PM »
Dear Zahra,

your question is not very specific. Of course in every program the implementation might differ a bit, but in general the programs should give the same accuracy using the same approximations

I assume you ask this question because you observe some unexpected deviations between old Gaussian results and your current Turbomole results. If you give us more information we can say a few words on it.

Which calculations your are doing?

If you are doing DFT calculations for example there can be several sources for differences between different programs. The integration grids might be different and this effect is not that small if you compare total energies, but should be insignificant if you look at energy differences. A more "famous" problem when comparing to Gaussian is the B3-LYP functional. Gaussian uses not the "official" variant, but a modified one. But how can use the Gaussian version in Turbomole as well.

Best wishes,
Gunnar

#### zahra

• Newbie
• Posts: 4
• Karma: +0/-0
##### Re: RI-approximation in Turbomole
« Reply #2 on: February 19, 2017, 08:56:11 PM »
Dear gunnar.schmitz;

Thanks a lot for your response. i am calculating the total energy at B-P86/def2-tzvpp with RI-j approximation. i want to replicate the results of Turbomole program by Gaussian09. the Bp86 functional is different in two software(Turbomole and Gaussian 09). Perdew 81 correlation function in default of Gaussian 09 for BP86functional is replaced with VWN(V) correlation function in Turbomole for this exchange correlation functional. i exerted this difference in gaussian09 and used the same basis set and the same auxiliary basis set. also, according to grid size m3 in Turbomole, i used int=(fine ) in Gaussian09 with the same 302 grid points , but results are different by about 2 hartree!!!

Best
Zahra

#### gunnar.schmitz

• Developers
• Jr. Member
• Posts: 14
• Karma: +0/-0
##### Re: RI-approximation in Turbomole
« Reply #3 on: February 19, 2017, 11:34:34 PM »
Dear Zahra,

okay, by the differences you now ensured that both programs use the same functional?

Anyway 2 Hartree are a large difference. Do you have some heavy elements in your system? I can imagine that one program uses ecps and the other one an all electron calculation.

Best wishes,
Gunnar

#### christof.haettig

• Global Moderator
• Sr. Member
• Posts: 265
• Karma: +0/-0
##### Re: RI-approximation in Turbomole
« Reply #4 on: March 21, 2017, 11:23:53 AM »
The comparison of total DFT energies between different codes is difficult, because the total energies are pretty sensitive to the integration grids
for the DFT functional and also sensitive to the details of XC functional. It would be better to compare energy differences.