Availability of MOLOPT basis sets

[Christopher O'Brien]

I am interested in obtaining additional MOLOPT basis sets for use with the GTH pseudopotentials. Specifically, I am investigating a mineral with Si, O, H, and Ca. The only Ca basis set for use with the GTH pseudo potential is of the DZVP-SR type. All of the other elements are using TZV2P basis sets. I understand that for Ca+/+2, a short range potential might be acceptable, but for consistently I would like to obtain the same basis set for all of the atoms in the system. 

Are there other Pseudopotential/basis pairs available that may be used with CP2K? 

[Christopher O'Brien]

Does anyone know how to generate an optimized basis set/pseudopotential for Ca (specifically of the TZV2P type)? I noticed that the ATOM module may permit me to do this by looking at the test files. Is there instructions for generating a basis/pseudo?  

[S Ling]

To generate a new pseudopotential for Ca, you may try the ATOMIC code, but I guess the current pseudopotential for Ca which was shipped with CP2K distribution should be good enough for you. What you need is to produce a new basis set for Ca, which is more complete than the DZVP-SR basis set provided by CP2K distribution. To do that, you can try with the OPTIMIZE_BASIS module. First, you perform a series of molecular calculations with a basis set which is close to the complete basis set limit. These accurate molecular calculations will serve as your reference. Then you choose a proper format of the basis set that you would like to optimize and provide initial guess for the Gaussian exponents and/or contraction coefficients, and you use the OPTIMIZE_BASIS module to optimize these variables by minimizing the electron density difference between the basis set that you would like to fit and your reference calculations. I am not sure if this explanation is clear enough. If not, I can try to produce a basis set for you, either using the same Gaussian exponents in DZVP-SR, or using more (and thus more diffused) Gaussian exponents if you believe your specific system will really benefit from that. Note the DZVP-SR basis set for Ca is already very diffused, with a small exponent of 0.055, which I think should be good enough for majority extended systems. Also keep in mind that a more complete basis set means much higher computational cost.

SL

[Matthias Krack]

Hi,

the attached file contains a set of Ca basis sets which I generated once still using the old basis set generator. These basis sets rather match the DZVP-GTH, TZVP-GTH, TZV2P-GTH basis set type sequence. Well, no guarantee that they will work for your system and thus a careful checking is mandatory as usual.

Matthias

[Christopher O'Brien]

S Ling,

I am very new to CP2K and have dabbled in Quantum Chemistry once a few years ago (NWCHEM). Usually I'm a MD/MC person, but when I'm working at the electronic structure level, I'm most comfortable with plane waves codes (VASP/Quantum Espresso). Consequently, I am a bit nervous about trying to generate my own basis set. I don't know enough about how they are constructed to feel comfortable performing a reference calculation or deciding on contraction coefficients. 

That being said, I hope that someone can enlighten me as to the meaning of the 'SR' option. Why should be basis be considered to be diffuse and still be called 'short-range'? Why does CP2K not have any non-short range basis sets for anything heavier than F? Will the DZVP-SR basis work well for Ca ions and Ca-O in clay?

Thanks for your responses,

Chris

[Christopher O'Brien]

Matthias,

I appreciate your generous donation of the basis set. 

[S Ling]

Hi

If you read the original paper on MOLOPT basis sets (see http://scitation.aip.org/content/aip/journal/jcp/127/11/10.1063/1.2770708), you should have some feeling about the localised basis sets used by CP2K. To some extent, the "short range" basis set which you are refering here is still long ranged, because it contains small Gaussian exponent. However, compared to the "long range" basis set which contain more and even smaller Gaussian exponents, it is indeed short ranged. So the "short range" and "long range" are based on comparisons. This is my understanding.

The DZVP-SR basis should be good enough for Ca ions and Ca-O in clay, but unfortunately, you don't have systematically better basis sets with which you can check whether your target properties are converged. The basis sets provided by Matthias are a good starting point for you. These basis sets can be further improved by optimising some of the contraction coefficients, which are set to 1.0 or 0.0 currently, but again, the optimisation of basis sets takes time. I will see whether I can find time to produce TZVP and maybe TZV2P quality basis sets for you.

SL

--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.
Visit this group at http://groups.google.com/group/cp2k.
For more options, visit https://groups.google.com/d/optout.

[S Ling]

Hi

I have attached two basis sets (TZVP + TZV2P) for Ca which I have produced for you. These basis sets share with same Gaussian exponents with the currently available basis set DZVP, but they contain more contraction coefficients (optimised in molecular calculations) and therefore should be systematically better in quality. You may do a series of calculations from DZVP to TZVP and to TZV2P, and check whether your target properties are converged. Keep in mind that the same quality of basis sets should be used for other elements in your system, i.e. when you use TZVP for Ca, you should also use TZVP for Si/O/H.

SL

[Feng Lai]

Hi

I want to produce the basis sets(TZVP + TZV2P) for copper as what you summarized in your slides. Firstly, i followed the example as you printed in your slides, but i  can't obtain the result as you got and printed in the next slide.Below are my some problems that confused me a lot:

1.I have no idea how to set the PP_BASIS section paramenters(S_EXPONENTS and P_EXPONENTS).

2.I got 40 exponents for every momentum in my result, and the maximum number of the exponent is 292075412.34116322. it is much higher than i have never seen before.Is the error that cause the unreasonable result.

3. And according to your computation result, I can't get the Na CBS as you printed the next slide.Is that the format problem?

1

2

3

myresult

在 2015年3月25日星期三 UTC+8下午6:11:20，S Ling写道：

[S Ling]

Hi

My slides serve as a guide, and not all details were included. If you run your own calculations following my slides, you should be able to find a new file "OPT_BASIS" in your directory, which contains the optimised exponents. The "Na CBS" basis set which you see on Slide 20 is not generated by the ATOM code. It was manually constructed, with optimised exponents from atomic calculations using the ATOM code. I am not sure how you got the 40 exponents in your output, but they seem like even-tempered basis set which CP2K uses to optimise pseudopotentials. I suspect your input is different from what I included in my slides.

To be short: I have optimised TZVP/TZV2P basis sets for majority of the elements in the periodic table (including Cu) which I hope to release in this summer. If you want to try the Cu basis set now, please send me an email.

SL

--

You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To post to this group, send email to cp...@googlegroups.com.

Visit this group at https://groups.google.com/group/cp2k.

[Feng Lai]

Hi

I would appreciate it if you could share the optimised TZVP/TZV2P basis sets with me. My email address is laifeng...@gmail.com.  And i am quite sure that my input is the same as yours, so i am very confused about how to get the paramenters in CBS. 
在 2016年6月30日星期四 UTC+8下午8:44:53，S Ling写道：