Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Hi Alberto, The manual of siesta-4.0.pdf did not address the issue on which pseudos should be used for vdW calculations. Are there practical details somewhere available or could you add some? Thanks in advance ticu On 17.02.2016 12:36, Alberto Garcia wrote: Hi, There is a subtle point concerning the use of the pseudopotential approximation with van der Waals functionals. In short, since the vdW xc term is non-local, the influence of the core can extend beyond the nominal "pseudopotential cutoff", leading to pseudos that can be pathological (for example, by converging to the coulomb tail too far out). For this reason, pseudopotentials generated with van der Waals functionals are not recommended. Instead, we recommend to generate the pseudopotentials with a good GGA functional like PBE, optionally changing the exchange part by that of the vdW functional to be used. This note will be added to the Siesta documentation as soon as possible. Best regards, Alberto On 17 Feb 2016, at 00:56, I. Campswrote: Hello Ticu, You have to generate your own vdW-DFT pseudos. What I did: - Download any INP (LDA or GGA) file from the SIESTA pseudos database for your atoms, - Change the head of the INP file and put the corresponding code for your desired flavor of vdW-DFT functional (the codes for each functional type and the place in the INP file you can find in the ATOM manual), - Generate the pseudo using the ATOM program (comes with SIESTA and should be compiled separately with the same arch.file), - Test the generated pseudo with the scripts that come with SIESTA. []'s, Camps On Tue, Feb 16, 2016 at 9:10 PM, ticu Cubot wrote: Hi Everybody! Because the manual of newly released 4.0b-485 does not address this point, I would like to ask you: are the ABINIT pseudopotentials suitable for vdW-DFT calculations and, if so, which flavour (_lda, _gga)? Many thanks, ticu
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Dear all Kindly suggest what to do out of two options written below %block XC.hybrid 2 GGA revPBE 1 0 VDW DRSLL 0 1 %endblock XC.hybrid Or this? XC.Functional VDW XC.authors DRSLL Sent from Samsung Mobile Original message From: "I. Camps"Date:11/08/2016 7:13 PM (GMT+05:30) To: siesta-l@uam.es Subject: Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations Dear Seyed, I have to agree with you. It is confusing. If you get the manual (both SIESTA and ATOM) and read some tutorials, that´s the path (change the INP file, then test using the scripts). With the email of Professor Alberto, I really do not now to match the two information. []'s, Camps On Thu, Aug 11, 2016 at 10:31 AM, Seyed Mohammad Tabatabaei wrote: Dear Professor Camps, I disagree with your statement and I refer to the message by Alberto Garcia: "For this reason, pseudopotentials generated with van der Waals functionals are not recommended. Instead, we recommend to generate the pseudopotentials with a good GGA functional like PBE, optionally changing the exchange part by that of the vdW functional to be used." Alberto is certain about using PBE pseudos. One can even download them from the SIESTA website. I just do not understand the meaning of "optionally changing the exchange part by that of the vdW functional to be used". Does it mean this? %block XC.hybrid 2 GGArevPBE 1 0 VDW DRSLL0 1 %endblock XC.hybrid Or this? XC.Functional VDW XC.authors DRSLL Bests, On Tue, Aug 9, 2016 at 5:52 PM, I. Camps wrote: Seyed, I think that you just have to generate the vdW pseudo directly using ATOM. Only that. Following the manual (ATOM 4.0.2 manual, page 13), you have the following flavors of vdW pseudos: Van der Waals (VDW) density functionals, implemented as Rom´an-P´erez and Soler, PRL 103, 096102 (2009): ∗ vw or vf: DRSLL (Dion et al, PRL 92, 246401 (2004)) ∗ vl: LMKLL (Lee et al, PRB 82, 081101 (2010)) ∗ vk: KBM (Klimes, Bowler, and Michaelides, JPCM 22, 022201 (2009)) ∗ vc: C09 (Cooper, PRB 81, 161104 (2010)) ∗ vb: BH (Berland and Hyldgaard, PRB 89, 035412 (2014)) ∗ vv: VV (Vydrov and VanVoorhis, JCP 133, 244103 (2010)) []'s, Camps On Tue, Aug 9, 2016 at 9:33 AM, Seyed Mohammad Tabatabaei wrote: Dear all, As I have understood, the DRSLL method for including the vdW interactions takes its correlation from the revPBE but the exchange part is innovated by the authors in the original DRSLL paper to include van der Waals interactions. Now, should I use PBE pseudos or should I use revPBE pseudos in my vdW calculations? Bests,
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Please guide me How to use these psf.in our fdf file On 11 Aug 2016 19:10, "Seyed Mohammad Tabatabaei"wrote: > Sorry all SIESTERS, > > I want to apologize and correct my previous comment. The following is > correct: > > The DRSLL method for including the vdW interactions takes its *exchange* > from the revPBE but the *correlation* part is innovated by the authors in > the original DRSLL paper to include van der Waals interactions. > > > > On Tue, Aug 9, 2016 at 5:03 PM, Seyed Mohammad Tabatabaei < > smt...@gmail.com> wrote: > >> Dear all, >> >> As I have understood, the DRSLL method for including the vdW interactions >> takes its correlation from the revPBE but the exchange part is innovated by >> the authors in the original DRSLL paper to include van der Waals >> interactions. Now, should I use PBE pseudos or should I use revPBE pseudos >> in my vdW calculations? >> >> Bests, >> > >
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Dear Seyed, I have to agree with you. It is confusing. If you get the manual (both SIESTA and ATOM) and read some tutorials, that´s the path (change the INP file, then test using the scripts). With the email of Professor Alberto, I really do not now to match the two information. []'s, Camps On Thu, Aug 11, 2016 at 10:31 AM, Seyed Mohammad Tabatabaei < smt...@gmail.com> wrote: > Dear Professor Camps, > > I disagree with your statement and I refer to the message by Alberto > Garcia: > > "For this reason, pseudopotentials generated with van der Waals > functionals are not recommended. Instead, we recommend to generate the > pseudopotentials with a good GGA functional like PBE, optionally changing > the exchange part by that of the vdW functional to be used." Alberto is > certain about using PBE pseudos. One can even download them from the SIESTA > website. > > I just do not understand the meaning of "optionally changing the exchange > part by that of the vdW functional to be used". Does it mean this? > > %block XC.hybrid > 2 > GGArevPBE 1 0 > VDW DRSLL0 1 > %endblock XC.hybrid > > Or this? > > XC.Functional VDW > XC.authors DRSLL > > Bests, > > > > On Tue, Aug 9, 2016 at 5:52 PM, I. Campswrote: > >> Seyed, >> >> I think that you just have to generate the vdW pseudo directly using >> ATOM. Only that. >> >> Following the manual (ATOM 4.0.2 manual, page 13), you have the following >> flavors of vdW pseudos: >> >> Van der Waals (VDW) density functionals, implemented as Rom´an-P´erez and >> Soler, >> PRL 103, 096102 (2009): >> ∗ vw or vf: DRSLL (Dion et al, PRL 92, 246401 (2004)) >> ∗ vl: LMKLL (Lee et al, PRB 82, 081101 (2010)) >> ∗ vk: KBM (Klimes, Bowler, and Michaelides, JPCM 22, 022201 (2009)) >> ∗ vc: C09 (Cooper, PRB 81, 161104 (2010)) >> ∗ vb: BH (Berland and Hyldgaard, PRB 89, 035412 (2014)) >> ∗ vv: VV (Vydrov and VanVoorhis, JCP 133, 244103 (2010)) >> >> []'s, >> >> Camps >> >> On Tue, Aug 9, 2016 at 9:33 AM, Seyed Mohammad Tabatabaei < >> smt...@gmail.com> wrote: >> >>> Dear all, >>> >>> As I have understood, the DRSLL method for including the vdW >>> interactions takes its correlation from the revPBE but the exchange part is >>> innovated by the authors in the original DRSLL paper to include van der >>> Waals interactions. Now, should I use PBE pseudos or should I use revPBE >>> pseudos in my vdW calculations? >>> >>> Bests, >>> >> >> >
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Dear Professor Camps, I disagree with your statement and I refer to the message by Alberto Garcia: "For this reason, pseudopotentials generated with van der Waals functionals are not recommended. Instead, we recommend to generate the pseudopotentials with a good GGA functional like PBE, optionally changing the exchange part by that of the vdW functional to be used." Alberto is certain about using PBE pseudos. One can even download them from the SIESTA website. I just do not understand the meaning of "optionally changing the exchange part by that of the vdW functional to be used". Does it mean this? %block XC.hybrid 2 GGArevPBE 1 0 VDW DRSLL0 1 %endblock XC.hybrid Or this? XC.Functional VDW XC.authors DRSLL Bests, On Tue, Aug 9, 2016 at 5:52 PM, I. Campswrote: > Seyed, > > I think that you just have to generate the vdW pseudo directly using ATOM. > Only that. > > Following the manual (ATOM 4.0.2 manual, page 13), you have the following > flavors of vdW pseudos: > > Van der Waals (VDW) density functionals, implemented as Rom´an-P´erez and > Soler, > PRL 103, 096102 (2009): > ∗ vw or vf: DRSLL (Dion et al, PRL 92, 246401 (2004)) > ∗ vl: LMKLL (Lee et al, PRB 82, 081101 (2010)) > ∗ vk: KBM (Klimes, Bowler, and Michaelides, JPCM 22, 022201 (2009)) > ∗ vc: C09 (Cooper, PRB 81, 161104 (2010)) > ∗ vb: BH (Berland and Hyldgaard, PRB 89, 035412 (2014)) > ∗ vv: VV (Vydrov and VanVoorhis, JCP 133, 244103 (2010)) > > []'s, > > Camps > > On Tue, Aug 9, 2016 at 9:33 AM, Seyed Mohammad Tabatabaei < > smt...@gmail.com> wrote: > >> Dear all, >> >> As I have understood, the DRSLL method for including the vdW interactions >> takes its correlation from the revPBE but the exchange part is innovated by >> the authors in the original DRSLL paper to include van der Waals >> interactions. Now, should I use PBE pseudos or should I use revPBE pseudos >> in my vdW calculations? >> >> Bests, >> > >
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Seyed, I think that you just have to generate the vdW pseudo directly using ATOM. Only that. Following the manual (ATOM 4.0.2 manual, page 13), you have the following flavors of vdW pseudos: Van der Waals (VDW) density functionals, implemented as Rom´an-P´erez and Soler, PRL 103, 096102 (2009): ∗ vw or vf: DRSLL (Dion et al, PRL 92, 246401 (2004)) ∗ vl: LMKLL (Lee et al, PRB 82, 081101 (2010)) ∗ vk: KBM (Klimes, Bowler, and Michaelides, JPCM 22, 022201 (2009)) ∗ vc: C09 (Cooper, PRB 81, 161104 (2010)) ∗ vb: BH (Berland and Hyldgaard, PRB 89, 035412 (2014)) ∗ vv: VV (Vydrov and VanVoorhis, JCP 133, 244103 (2010)) []'s, Camps On Tue, Aug 9, 2016 at 9:33 AM, Seyed Mohammad Tabatabaeiwrote: > Dear all, > > As I have understood, the DRSLL method for including the vdW interactions > takes its correlation from the revPBE but the exchange part is innovated by > the authors in the original DRSLL paper to include van der Waals > interactions. Now, should I use PBE pseudos or should I use revPBE pseudos > in my vdW calculations? > > Bests, >
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Dear all, As I have understood, the DRSLL method for including the vdW interactions takes its correlation from the revPBE but the exchange part is innovated by the authors in the original DRSLL paper to include van der Waals interactions. Now, should I use PBE pseudos or should I use revPBE pseudos in my vdW calculations? Bests,
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Unfortunately, I can not find any connection between my question and the answer. On 8/5/16, Peter Kovalwrote: > In case of silver atom, the pseudos provided by Rivero etal give a > screwed up TDDFT spectra (without plasmonic peak, which is easy to > reproduce with PPs provided by icmab > http://departments.icmab.es/leem/siesta/Databases/Pseudopotentials/periodictable-intro.html) > > http://stacks.iop.org/0953-8984/28/i=21/a=214001 > > Peter > > > On Wed, Feb 17, 2016 at 1:00 PM, ticu Cubot wrote: >> Thank you very much for your reactions. Alberto's mail (indirectly) >> implies >> that GGA-PBE pseudos should be "reasonable". Along this line, can I >> (without >> experience in generating own pseudos, as Camps kindly suggested) conclude >> that and use PBE pseudos from ABINIT or the SI of the recent study >> >> http://www.sciencedirect.com/science/article/pii/S0927025614007940 >> >> Thanks again, >> >> ticu >> >> On Wed, Feb 17, 2016 at 12:36 PM, Alberto Garcia >> wrote: >>> >>> Hi, >>> >>> There is a subtle point concerning the use of the pseudopotential >>> approximation with van der Waals functionals. In short, since the vdW xc >>> term >>> is non-local, the influence of the core can extend beyond the nominal >>> "pseudopotential cutoff", leading to pseudos that can be pathological >>> (for >>> example, by converging to the coulomb tail too far out). >>> >>> For this reason, pseudopotentials generated with van der Waals >>> functionals >>> are not recommended. Instead, we recommend to generate the >>> pseudopotentials >>> with a good GGA functional like PBE, optionally changing the exchange >>> part >>> by that of the vdW functional to be used. >>> >>> This note will be added to the Siesta documentation as soon as possible. >>> >>> Best regards, >>> >>> Alberto >>> >>> >>> On 17 Feb 2016, at 00:56, I. Camps wrote: >>> >>> > Hello Ticu, >>> > >>> > You have to generate your own vdW-DFT pseudos. >>> > >>> > What I did: >>> > - Download any INP (LDA or GGA) file from the SIESTA pseudos database >>> > for your atoms, >>> > - Change the head of the INP file and put the corresponding code for >>> > your desired flavor of vdW-DFT functional (the codes for each >>> > functional >>> > type and the place in the INP file you can find in the ATOM manual), >>> > - Generate the pseudo using the ATOM program (comes with SIESTA and >>> > should be compiled separately with the same arch.file), >>> > - Test the generated pseudo with the scripts that come with SIESTA. >>> > >>> > >>> > []'s, >>> > >>> > Camps >>> > >>> > On Tue, Feb 16, 2016 at 9:10 PM, ticu Cubot >>> > wrote: >>> > Hi Everybody! >>> > >>> > Because the manual of newly released 4.0b-485 does not address this >>> > point, I would like to ask you: are the ABINIT pseudopotentials >>> > suitable for >>> > vdW-DFT calculations and, if so, which flavour (_lda, _gga)? >>> > >>> > Many thanks, >>> > ticu >>> > >>> >> >
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
In case of silver atom, the pseudos provided by Rivero etal give a screwed up TDDFT spectra (without plasmonic peak, which is easy to reproduce with PPs provided by icmab http://departments.icmab.es/leem/siesta/Databases/Pseudopotentials/periodictable-intro.html) http://stacks.iop.org/0953-8984/28/i=21/a=214001 Peter On Wed, Feb 17, 2016 at 1:00 PM, ticu Cubotwrote: > Thank you very much for your reactions. Alberto's mail (indirectly) implies > that GGA-PBE pseudos should be "reasonable". Along this line, can I (without > experience in generating own pseudos, as Camps kindly suggested) conclude > that and use PBE pseudos from ABINIT or the SI of the recent study > > http://www.sciencedirect.com/science/article/pii/S0927025614007940 > > Thanks again, > > ticu > > On Wed, Feb 17, 2016 at 12:36 PM, Alberto Garcia wrote: >> >> Hi, >> >> There is a subtle point concerning the use of the pseudopotential >> approximation with van der Waals functionals. In short, since the vdW xc >> term >> is non-local, the influence of the core can extend beyond the nominal >> "pseudopotential cutoff", leading to pseudos that can be pathological (for >> example, by converging to the coulomb tail too far out). >> >> For this reason, pseudopotentials generated with van der Waals functionals >> are not recommended. Instead, we recommend to generate the pseudopotentials >> with a good GGA functional like PBE, optionally changing the exchange part >> by that of the vdW functional to be used. >> >> This note will be added to the Siesta documentation as soon as possible. >> >> Best regards, >> >> Alberto >> >> >> On 17 Feb 2016, at 00:56, I. Camps wrote: >> >> > Hello Ticu, >> > >> > You have to generate your own vdW-DFT pseudos. >> > >> > What I did: >> > - Download any INP (LDA or GGA) file from the SIESTA pseudos database >> > for your atoms, >> > - Change the head of the INP file and put the corresponding code for >> > your desired flavor of vdW-DFT functional (the codes for each functional >> > type and the place in the INP file you can find in the ATOM manual), >> > - Generate the pseudo using the ATOM program (comes with SIESTA and >> > should be compiled separately with the same arch.file), >> > - Test the generated pseudo with the scripts that come with SIESTA. >> > >> > >> > []'s, >> > >> > Camps >> > >> > On Tue, Feb 16, 2016 at 9:10 PM, ticu Cubot wrote: >> > Hi Everybody! >> > >> > Because the manual of newly released 4.0b-485 does not address this >> > point, I would like to ask you: are the ABINIT pseudopotentials suitable >> > for >> > vdW-DFT calculations and, if so, which flavour (_lda, _gga)? >> > >> > Many thanks, >> > ticu >> > >> >
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Dear Alberto Garcia, I would be really grateful if you clarify a little more about the following for me: "Instead, we recommend to generate the pseudopotentials with a good GGA functional like PBE, optionally changing the exchange part by that of the vdW functional to be used." Given the fact that .inp files are used with ATOM to generate pseudos, do you mean that I have to change the .inp files of my previous GGA-PBE pseudos from "pb" to "vw" (Not Changing Anything Else) and generate my vdW pseudos? Or simply use pseudos generated with the "pb" option in the .inp file (GGA-PBE pseudos) and set the following lines in the main .fdf file instead? XC.Functional VDW XC.authors DRSLL If I use the latter strategy, I think the following warning in the output file is always inevitable: xc_check: Exchange-correlation functional: xc_check: VDW Dion-Rydberg-Schroeder-Langreth-Lundqvist xc_check: WARNING: Pseudopotential generated with GGA PBE functional Please guide me on this issue. Should I ignore this warning? Bests, Mohammad,
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Thank you very much for your reactions. Alberto's mail (indirectly) implies that GGA-PBE pseudos should be "reasonable". Along this line, can I (without experience in generating own pseudos, as Camps kindly suggested) conclude that and use PBE pseudos from ABINIT or the SI of the recent study http://www.sciencedirect.com/science/article/pii/S0927025614007940 Thanks again, ticu On Wed, Feb 17, 2016 at 12:36 PM, Alberto Garciawrote: > Hi, > > There is a subtle point concerning the use of the pseudopotential > approximation with van der Waals functionals. In short, since the vdW xc > term > is non-local, the influence of the core can extend beyond the nominal > "pseudopotential cutoff", leading to pseudos that can be pathological (for > example, by converging to the coulomb tail too far out). > > For this reason, pseudopotentials generated with van der Waals functionals > are not recommended. Instead, we recommend to generate the pseudopotentials > with a good GGA functional like PBE, optionally changing the exchange part > by that of the vdW functional to be used. > > This note will be added to the Siesta documentation as soon as possible. > > Best regards, > > Alberto > > > On 17 Feb 2016, at 00:56, I. Camps wrote: > > > Hello Ticu, > > > > You have to generate your own vdW-DFT pseudos. > > > > What I did: > > - Download any INP (LDA or GGA) file from the SIESTA pseudos database > for your atoms, > > - Change the head of the INP file and put the corresponding code for > your desired flavor of vdW-DFT functional (the codes for each functional > type and the place in the INP file you can find in the ATOM manual), > > - Generate the pseudo using the ATOM program (comes with SIESTA and > should be compiled separately with the same arch.file), > > - Test the generated pseudo with the scripts that come with SIESTA. > > > > > > []'s, > > > > Camps > > > > On Tue, Feb 16, 2016 at 9:10 PM, ticu Cubot wrote: > > Hi Everybody! > > > > Because the manual of newly released 4.0b-485 does not address this > point, I would like to ask you: are the ABINIT pseudopotentials suitable > for vdW-DFT calculations and, if so, which flavour (_lda, _gga)? > > > > Many thanks, > > ticu > > > >
Re: [SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Dear Alberto, Thank you very much for your clarifications. How can I change the exchange part by that of the vdW functional? []'s, Camps On Wed, Feb 17, 2016 at 9:36 AM, Alberto Garciawrote: > Hi, > > There is a subtle point concerning the use of the pseudopotential > approximation with van der Waals functionals. In short, since the vdW xc > term > is non-local, the influence of the core can extend beyond the nominal > "pseudopotential cutoff", leading to pseudos that can be pathological (for > example, by converging to the coulomb tail too far out). > > For this reason, pseudopotentials generated with van der Waals functionals > are not recommended. Instead, we recommend to generate the pseudopotentials > with a good GGA functional like PBE, optionally changing the exchange part > by that of the vdW functional to be used. > > This note will be added to the Siesta documentation as soon as possible. > > Best regards, > > Alberto > > > On 17 Feb 2016, at 00:56, I. Camps wrote: > > > Hello Ticu, > > > > You have to generate your own vdW-DFT pseudos. > > > > What I did: > > - Download any INP (LDA or GGA) file from the SIESTA pseudos database > for your atoms, > > - Change the head of the INP file and put the corresponding code for > your desired flavor of vdW-DFT functional (the codes for each functional > type and the place in the INP file you can find in the ATOM manual), > > - Generate the pseudo using the ATOM program (comes with SIESTA and > should be compiled separately with the same arch.file), > > - Test the generated pseudo with the scripts that come with SIESTA. > > > > > > []'s, > > > > Camps > > > > On Tue, Feb 16, 2016 at 9:10 PM, ticu Cubot wrote: > > Hi Everybody! > > > > Because the manual of newly released 4.0b-485 does not address this > point, I would like to ask you: are the ABINIT pseudopotentials suitable > for vdW-DFT calculations and, if so, which flavour (_lda, _gga)? > > > > Many thanks, > > ticu > > > >
[SIESTA-L] A note of warning regarding pseudopotentials for vdW-DFT calculations
Hi, There is a subtle point concerning the use of the pseudopotential approximation with van der Waals functionals. In short, since the vdW xc term is non-local, the influence of the core can extend beyond the nominal "pseudopotential cutoff", leading to pseudos that can be pathological (for example, by converging to the coulomb tail too far out). For this reason, pseudopotentials generated with van der Waals functionals are not recommended. Instead, we recommend to generate the pseudopotentials with a good GGA functional like PBE, optionally changing the exchange part by that of the vdW functional to be used. This note will be added to the Siesta documentation as soon as possible. Best regards, Alberto On 17 Feb 2016, at 00:56, I. Campswrote: > Hello Ticu, > > You have to generate your own vdW-DFT pseudos. > > What I did: > - Download any INP (LDA or GGA) file from the SIESTA pseudos database for > your atoms, > - Change the head of the INP file and put the corresponding code for your > desired flavor of vdW-DFT functional (the codes for each functional type and > the place in the INP file you can find in the ATOM manual), > - Generate the pseudo using the ATOM program (comes with SIESTA and should be > compiled separately with the same arch.file), > - Test the generated pseudo with the scripts that come with SIESTA. > > > []'s, > > Camps > > On Tue, Feb 16, 2016 at 9:10 PM, ticu Cubot wrote: > Hi Everybody! > > Because the manual of newly released 4.0b-485 does not address this point, I > would like to ask you: are the ABINIT pseudopotentials suitable for vdW-DFT > calculations and, if so, which flavour (_lda, _gga)? > > Many thanks, > ticu >