Band gap-HSE06-UiO-66

[jingyu...@gmail.com]

Hello everyone,

I am trying to compare with HSE06 functional using VASP and CP2K. I found out that the band gap calculated using VASP is about 3.7 eV for UiO-66 which is very close to the experimental value 3.6-3.7 eV. 

However, when I use CP2K, the band gap is 4.07 eV, which is almost the same with the data reported by the following paper.  (Understanding Intrinsic Light Absorption Properties of UiO-66 Frameworks: A Combined Theoretical and Experimental Study, Inorg. Chem. 2015, 54, 22, 10701–10710) 

I wonder if the developer could comment on the accuracy? Or any room to improve the accuracy.  (HSE06 functional section I used is the default that posted on the CP2K website or lectures by Dr. Sanliang Ling. ) 

Another question: Only shortrange keyword is allowed to use for HSE06? 

Thank you, 

Jenny 

[Lucas Lodeiro]

Hello Jenny,

AFAIK HSE functional is shortrange functional by definition... it is constructed using the erf(wr) technique and leaving only the short range term. WRT the value of the band gap, the band gap is very dependent on the reciprocal space sampling, whether the band gap is not vertical in reciprocal space or either the eigenvalues are highly dependent on the sampling, so I guess you use kpoints in VASP, did you use the same structure/cell and kpoints in CP2K instead? This is very important to compare both results.

Also, the accuracy must be set by your calculation settings, as EPS_DEFAULT, EPS_PGF_ORB, your basis set and auxiliary basis set (ADMM). Also the possible screening for the HF part, and the cutoff, which must be sufficient long to ensure the short range decay to low values...

Regards - Lucas

--
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 view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/7c9518cd-8f3c-4d1e-acb8-0b89caed128dn%40googlegroups.com.

[Jingyun Ye]

Hello Lucas, 

Thank you for sharing your thoughts. So normally for the MOFs system, either VASP or CP2K, gama point is usually selected for calculation since the unit cell has several hundreds of atoms. For the test materials, it only has the direct band gap. 

The xyz coordinate, lattice constants are the same for both VASP and CP2K tests. 

I don't think it is because of  EPS_DEFAULT, EPS_PGF_ORB because they are in the magnitudes of -14, and -60. 

I used the basis set: DZVP and auxiliary basis set (ADMM): a combination of c-FIT11 and cFIT3, which are the maximum I can afford, or else, the calculation will run out of memory. I think this might be the case for the large systems.  

Could you please share more about the CUTOFF? You mean the GHT basis set (400) and plane wave cutoff (60), right? I think my cutoff is reasonable since I test the convergence for energy but for PBE functional. 

I think this will be a very helpful discussion for all. 

Thank you,

Jenny 

You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/KgKHoHhouyc/unsubscribe.
To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PJUcj-bNDGAfo49wc7wMLy9E5KK2-mOTPJ%2ByPnH%3D4psqQ%40mail.gmail.com.

[Matt Watkins]

Hi,

I'd try running with a bigger primary basis set (TZVP) and see if the band gap is sensitive to this.

Matt

[Jürg Hutter]

Hi

I assume the difference is due to basis sets. Plane waves usually describe conduction band states better than localized orbitals. So increase the primary basis.

Important:
if you are using a new version of CP2K, you should not set EPS_PGF_ORB.

-> https://www.cp2k.org/faq:hfx_eps_warning
Since version 2023.2 The occupancy of the Kohn-Sham matrix depends on whether atomic pairs are neighbours. Two atoms are considered neighbours if their basis functions (Gaussians) are overlapping, which is controlled through the EPS_PFG_ORB keyword. In large cells, the KS matrix can be sparse because atomic pairs can be further apart. Setting the MIN_PAIR_LIST_RADIUS to -1 makes sure all atomic pairs within the unit cell are considered neighbours, therefore ensuring a full KS matrix. This is considerably more efficient than reducing EPS_PFG_ORB to unreasonably small numbers.

regards
JH

________________________________________
From: cp...@googlegroups.com <cp...@googlegroups.com> on behalf of Jingyun Ye <jingyu...@gmail.com>
Sent: Wednesday, October 4, 2023 3:30 AM
To: cp...@googlegroups.com
Subject: Re: [CP2K:19288] Band gap-HSE06-UiO-66

Hello Lucas,

Thank you for sharing your thoughts. So normally for the MOFs system, either VASP or CP2K, gama point is usually selected for calculation since the unit cell has several hundreds of atoms. For the test materials, it only has the direct band gap.

The xyz coordinate, lattice constants are the same for both VASP and CP2K tests.

I don't think it is because of EPS_DEFAULT, EPS_PGF_ORB because they are in the magnitudes of -14, and -60.

I used the basis set: DZVP and auxiliary basis set (ADMM): a combination of c-FIT11 and cFIT3, which are the maximum I can afford, or else, the calculation will run out of memory. I think this might be the case for the large systems.

Could you please share more about the CUTOFF? You mean the GHT basis set (400) and plane wave cutoff (60), right? I think my cutoff is reasonable since I test the convergence for energy but for PBE functional.

I think this will be a very helpful discussion for all.

Thank you,
Jenny





On Tue, Oct 3, 2023 at 6:48 PM Lucas Lodeiro <eluni...@gmail.com<mailto:eluni...@gmail.com>> wrote:
Hello Jenny,

AFAIK HSE functional is shortrange functional by definition... it is constructed using the erf(wr) technique and leaving only the short range term. WRT the value of the band gap, the band gap is very dependent on the reciprocal space sampling, whether the band gap is not vertical in reciprocal space or either the eigenvalues are highly dependent on the sampling, so I guess you use kpoints in VASP, did you use the same structure/cell and kpoints in CP2K instead? This is very important to compare both results.
Also, the accuracy must be set by your calculation settings, as EPS_DEFAULT, EPS_PGF_ORB, your basis set and auxiliary basis set (ADMM). Also the possible screening for the HF part, and the cutoff, which must be sufficient long to ensure the short range decay to low values...
Regards - Lucas

El mar, 3 oct 2023 a las 17:37, jingyu...@gmail.com<mailto:jingyu...@gmail.com> (<jingyu...@gmail.com<mailto:jingyu...@gmail.com>>) escribió:
Hello everyone,

I am trying to compare with HSE06 functional using VASP and CP2K. I found out that the band gap calculated using VASP is about 3.7 eV for UiO-66 which is very close to the experimental value 3.6-3.7 eV.

However, when I use CP2K, the band gap is 4.07 eV, which is almost the same with the data reported by the following paper. (Understanding Intrinsic Light Absorption Properties of UiO-66 Frameworks: A Combined Theoretical and Experimental Study, Inorg. Chem. 2015, 54, 22, 10701–10710)

I wonder if the developer could comment on the accuracy? Or any room to improve the accuracy. (HSE06 functional section I used is the default that posted on the CP2K website or lectures by Dr. Sanliang Ling. )

Another question: Only shortrange keyword is allowed to use for HSE06?

Thank you,
Jenny

--
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<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/7c9518cd-8f3c-4d1e-acb8-0b89caed128dn%40googlegroups.com<https://groups.google.com/d/msgid/cp2k/7c9518cd-8f3c-4d1e-acb8-0b89caed128dn%40googlegroups.com?utm_medium=email&utm_source=footer>.

--
You received this message because you are subscribed to a topic in the Google Groups "cp2k" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/cp2k/KgKHoHhouyc/unsubscribe.

To unsubscribe from this group and all its topics, send an email to cp2k+uns...@googlegroups.com<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CAOFT4PJUcj-bNDGAfo49wc7wMLy9E5KK2-mOTPJ%2ByPnH%3D4psqQ%40mail.gmail.com<https://groups.google.com/d/msgid/cp2k/CAOFT4PJUcj-bNDGAfo49wc7wMLy9E5KK2-mOTPJ%2ByPnH%3D4psqQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.

--
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<mailto:cp2k+uns...@googlegroups.com>.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/CALf-S3QzfO3wMpHnXn2Km4Ex8wwt-WpEMrSg2Xdt%3DPu482z9sQ%40mail.gmail.com<https://groups.google.com/d/msgid/cp2k/CALf-S3QzfO3wMpHnXn2Km4Ex8wwt-WpEMrSg2Xdt%3DPu482z9sQ%40mail.gmail.com?utm_medium=email&utm_source=footer>.