From owner-chemistry@ccl.net Mon Jan 21 13:16:00 2019
From: "Susi Lehtola susi.lehtola|a|alumni.helsinki.fi"
To: CCL
Subject: CCL:G: All electron basis set for H2Te
Message-Id: <-53592-190121125727-9840-nvsGz15FNA/ho9APjf5sUw*|*server.ccl.net>
X-Original-From: Susi Lehtola
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Content-Type: text/plain; charset=utf-8; format=flowed
Date: Mon, 21 Jan 2019 19:57:18 +0200
MIME-Version: 1.0
Sent to CCL by: Susi Lehtola [susi.lehtola%x%alumni.helsinki.fi]
On 1/21/19 2:27 PM, Mo Fateh mo.fateh+/-yahoo.com wrote:
>
> Sent to CCL by: "Mo Fateh" [mo.fateh]_[yahoo.com] Dear CCL
> Subscribers,
>
> I am going to do geometry optimization for H2O, H2S, H2Se, and H2Te
> using Gaussian 09. What is the suitable All electron basis set for
> optimization with B3LYP?
>
> I have checked the Gaussian website and found that the following bais
> sets are suitable:
>
> 1- DGDZVP basis set 2- QZVPand Def2 3- UGBS basis set
>
> Which is one you recommended? and Why? Please suggest any other
> suitable basis set
First, do you really need an all-electron treatment? Using an effective
core potential (ECP) typically gives you good results with a low cost.
The def2 sets are not all-electron, but use relativistic core potentials
starting from the 5th period.
In addition to def2, you can also use the correlation consistent series,
cc-pVXZ-PP, for the heavy atoms, which employs ECPs. These are probably
not in the code you are using, but you can download the sets from the
Basis Set Exchange (BSE) at https://bse.pnl.gov/bse/portal.
If you want to do all-electron calculations, then you have to include
relativistic effects to get any kind of reasonable accuracy. It looks
like Gaussian can do second-order Douglas-Kroll-Hess (DKH2)
calculations, and that the program uses a finite nuclear model in such a
case. But, when you do a relativistic calculation, you have to make sure
that the basis set has actually been designed for the method you are using.
Unfortunately, there aren't many relativistic basis sets available on
BSE. IIRC, the ANO-RCC set by AlmlĂ¶f, Roos, and coworkers available on
BSE has been parametrized with DKH2, so that would be an okay choice.
Then, there are the x2c-SV(P)all, x2c-SVPall, x2c-TZVPall and
x2c-TZVPPall basis sets by Pollak and Weigend, which are all-electron
versions of the def2 basis sets; however, these basis sets have been
formed with the eponymous X2C approach which differs from DKH2, and so
the contractions may not be right for DKH2.
If you want to play on the safe side, you could decontract these sets;
like I implied before, the main problem with different relativistic
models is that the core orbitals may be dissimilar. Using a fully
uncontracted basis set takes away this problem; of course, then your
basis set becomes larger.
Personally, I would recommend either using ECP sets like def2, or if you
really need all electrons, then I'd go with the x2c series. These are
both commonly used, so I'd expect fewer problems.
UGBS is a fine basis set for mean-field calculations on atoms. The
problem is just that it is pretty huge (71, 84, 163, and 179 for O, S,
Se, and Te, respectively), while it lacks polarization and correlation
functions altogether. This means that calculations in molecules and/or
at a post-HF level of theory will be unbalanced; you could trade off a
small decrease in accuracy for atoms for a large improvement for
molecules with the same number of basis functions by using a different
kind of basis set.
There are variants of UGBS with polarization functions, but
1) these are much larger than even the plain UGBS basis set; several
hundred functions per atom,
2) I haven't been able to find any description on how the polarized
basis sets have been formed, and
3) I am not aware of any accuracy benchmarks with these sets.
--
------------------------------------------------------------------
Mr. Susi Lehtola, PhD Junior Fellow, Adjunct Professor
susi.lehtola-*-alumni.helsinki.fi University of Helsinki
http://susilehtola.github.io/ Finland
------------------------------------------------------------------
Susi Lehtola, dosentti, FT tutkijatohtori
susi.lehtola-*-alumni.helsinki.fi Helsingin yliopisto
http://susilehtola.github.io/
------------------------------------------------------------------