[Q-e-developers] SCAN functional

Hsin-Yu Ko kosinyj at gmail.com
Mon Mar 27 04:27:37 CEST 2017 

Hello Paolo, and Davide (cc: Marcos),

We (Marcos and myself) are very interested in merging the SCAN 
implementation in to the main trunk and have got the permission from our 
PIs. At this point our implementation is based on roughly v5.1.0, and 
there has been considerable changes to the trunk from where we stand; we 
would like to ask for your advice on the best way to merge the code.

We could send you our version and/or try to perform what we have done on 
the main trunk. Please let us know your advice.

Thank you very much and best wishes,
Hsin-Yu

On 03/24/17 11:04, Vigil-Fowler, Derek wrote:
> Hi Paolo and Davide,
>      Many thanks for all your useful comments.
>
>      Paolo, thanks for letting me know about the problems with metagga and
> libxc in QE. For metagga¹s, do you know if any other codes have had any
> luck with dealing with the instability? I know the papers on SCAN that are
> out there now were done in VASP. Are there any papers from Kresse, et al.
> or others on taming the instability, or is that a deeply guarded secret?
> Or do they just run on systems where they don¹t run into an instability
> and give up on one¹s where there is one? I¹m just trying to understand the
> state of the art.
>
>       Davide, I appreciate your patch - it sounds very useful. Below my
> sign off is information from the graduate student who work on SCAN in QE
> with Roberto Car, Hsin-Yu Kuo (he worked on it with Marcos Andrade). They
> seemed to have used libxc, so we¹ll see what exactly that looks like when
> they get the ok to share the code. So, I¹d save yourself the work on SCAN
> for now. I will update this string as I hear back from Hsin-Yu.
>
>       Thanks.
>
> -Derek
>
> Thanks a lot for your kind email and interest in our implementation. Yes,
> we are very happy to share the code, and are also planning to check it in
> the QE trunk recently (we have not had time to merge the changes since we
> did the implementation based on version QE-5.1.2, which is quite different
> from the main trunk now...).
>
> At this point, our implementation utilizes libxc along with a few bug fix
> to the QE metagga routines for the CP module; on the other hand, for PW
> module, I believe that we also have a working version for SCF calculation
> thanks to Marcos (cc'd). While we still have to confirm with our
> PIs---Roberto Car and Annabelle Selloni, I believe they would be very
> happy to share. I will keep you posted.
>
>
>
> On 3/24/17, 5:00 AM, "q-e-developers-bounces at qe-forge.org on behalf of
> q-e-developers-request at qe-forge.org" <q-e-developers-bounces at qe-forge.org
> on behalf of q-e-developers-request at qe-forge.org> wrote:
>
>> Send Q-e-developers mailing list submissions to
>> 	q-e-developers at qe-forge.org
>>
>> To subscribe or unsubscribe via the World Wide Web, visit
>> 	http://qe-forge.org/mailman/listinfo/q-e-developers
>> or, via email, send a message with subject or body 'help' to
>> 	q-e-developers-request at qe-forge.org
>>
>> You can reach the person managing the list at
>> 	q-e-developers-owner at qe-forge.org
>>
>> When replying, please edit your Subject line so it is more specific
>> than "Re: Contents of Q-e-developers digest..."
>>
>>
>> Today's Topics:
>>
>>   1. Re: SCAN functional (Paolo Giannozzi)
>>   2. Re: Personal QE mirror (the "official" one is not working
>>      anymore) (Paolo Giannozzi)
>>   3. Re: SCAN functional (Vigil-Fowler, Derek)
>>   4. R: Re:  SCAN functional (Davide Ceresoli)
>>   5. Trouble with vc-relax: "Too many g vector" error
>>      (Mohammadreza Karamad)
>>
>>
>> ----------------------------------------------------------------------
>>
>> Message: 1
>> Date: Thu, 23 Mar 2017 14:31:56 +0100
>> From: Paolo Giannozzi <p.giannozzi at gmail.com>
>> Subject: Re: [Q-e-developers] SCAN functional
>> To: Davide Ceresoli <davide.ceresoli at cnr.it>,	General discussion list
>> 	for Quantum ESPRESSO developers	<q-e-developers at qe-forge.org>
>> Message-ID:
>> 	<CAPMgbCu5COh89OCkH35Cn33-AxVV-VsLXVDQQTuo0VM0OKu7gA at mail.gmail.com>
>> Content-Type: text/plain; charset=UTF-8
>>
>> Davide, it's not that simple. Meta-GGA suffers from numerical
>> instabilities. Until we figure out how to cope with it, meta-GGA in QE
>> will be almost unusable. The support of libxc in QE is limited and
>> hard to use. A better support would need a refactoring of the entire
>> XC machinery (which would be good anyway).
>>
>> I am personally interested in settling these issues once and for all
>> but it takes time and effort.
>>
>> Paolo
>>
>> On Thu, Mar 23, 2017 at 11:20 AM, Davide Ceresoli
>> <davide.ceresoli at cnr.it> wrote:
>>> Hi all,
>>>      if I may add my 2cents, the SCAN functional is implemented in
>>> LibXC.
>>> QE already uses LibXC to provide TPSS and TB09. No need to reinvent
>>> the wheel. I'd like to point out, on passing, that meta-GGA support
>>> in QE, somehow degraded with time. It should be revamped.
>>>
>>> Best wishes,
>>>      Davide
>>>
>>>
>>>
>>> On 03/23/2017 12:28 AM, Vigil-Fowler, Derek wrote:
>>>> Hi QE developers,
>>>>         My name is Derek Vigil-Fowler and I am a Director's
>>>> Postdoctoral Fellow at NREL in the US.
>>>>
>>>>         I am intersted in using the SCAN functional and saw some
>>>> activity in this forum back in October, but there was no resolution.
>>>> Paolo said something about hearing second hand about an implementation
>>>> in CP and someone else told me that Roberto Car's group had an
>>>> implementation, but I don't see anything in CP other than TPSS. I
>>>> thought I'd check back in on this since a decent amount of time has
>>>> elapsed. If you could let me know if there's an implementation
>>>> somewhere in QE community, that would be much appreciated. If not, I'll
>>>> plow ahead with an implementation.
>>>>
>>>>         Thanks for your time and attention.
>>>>
>>>> -Derek
>>>>
>>>> _______________________________________________
>>>> Q-e-developers mailing list
>>>> Q-e-developers at qe-forge.org
>>>> http://qe-forge.org/mailman/listinfo/q-e-developers
>>>>
>>> _______________________________________________
>>> Q-e-developers mailing list
>>> Q-e-developers at qe-forge.org
>>> http://qe-forge.org/mailman/listinfo/q-e-developers
>>
>>
>>
>> --
>> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
>> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
>> Phone +39-0432-558216, fax +39-0432-558222
>>
>>
>>
>> ------------------------------
>>
>> Message: 2
>> Date: Thu, 23 Mar 2017 14:37:54 +0100
>> From: Paolo Giannozzi <p.giannozzi at gmail.com>
>> Subject: Re: [Q-e-developers] Personal QE mirror (the "official" one
>> 	is not working anymore)
>> To: General discussion list for Quantum ESPRESSO developers
>> 	<q-e-developers at qe-forge.org>
>> Message-ID:
>> 	<CAPMgbCvvuGbip4dGb7R+PC_+ehV8hkJ+GYLwLhOURWLew-gxnQ at mail.gmail.com>
>> Content-Type: text/plain; charset=UTF-8
>>
>> Anton, this is clear to everybody, we just have to decide the details
>> on how and where to host the repository.
>>
>> Paolo
>>
>> On Thu, Mar 23, 2017 at 10:17 AM, Anton Kozhevnikov
>> <a.v.kozhevnikov at gmail.com> wrote:
>>> Just to promote the further Git discussion:
>>> Git is a preferred revision control system for CSCS (we use GitHub and
>>> GitLab heavily). EPFL AiiDA team is also using Git and I know that
>>> CINECA
>>> guys are also in favour of Git. I can safely assume that the people
>> >from the
>>> US national labs who are involved in the QE development would also
>>> prefer to
>>> deal with Git.
>>>
>>>
>>> _______________________________________________
>>> Q-e-developers mailing list
>>> Q-e-developers at qe-forge.org
>>> http://qe-forge.org/mailman/listinfo/q-e-developers
>>>
>>
>>
>>
>> --
>> Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
>> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
>> Phone +39-0432-558216, fax +39-0432-558222
>>
>>
>> ------------------------------
>>
>> Message: 3
>> Date: Thu, 23 Mar 2017 20:23:21 +0000
>> From: "Vigil-Fowler, Derek" <Derek.Vigil-Fowler at nrel.gov>
>> Subject: Re: [Q-e-developers] SCAN functional
>> To: "q-e-developers at qe-forge.org" <q-e-developers at qe-forge.org>
>> Message-ID: <D4F98AA2.568D%derek.vigil-fowler at nrel.gov>
>> Content-Type: text/plain; charset="iso-8859-1"
>>
>> Hi Davide,
>>      Many thanks for letting me know that SCAN is in LibXC. I will try
>> that first. I am sorry to hear that meta-GGAs are not working well in QE.
>> I hope it does not cause problem for intended calculations. I?ll certainly
>> report any problems I have.
>>
>> -Derek
>>
>> On 3/23/17, 5:00 AM, "q-e-developers-bounces at qe-forge.org on behalf of
>> q-e-developers-request at qe-forge.org" <q-e-developers-bounces at qe-forge.org
>> on behalf of q-e-developers-request at qe-forge.org> wrote:
>>
>>> Re: SCAN functional
>>
>>
>>
>>
>> ------------------------------
>>
>> Message: 4
>> Date: Thu, 23 Mar 2017 21:46:43 +0100
>> From: Davide Ceresoli <davide.ceresoli at cnr.it>
>> Subject: [Q-e-developers] R: Re:  SCAN functional
>> To: General discussion list for Quantum ESPRESSO developers
>> 	<q-e-developers at qe-forge.org>
>> Message-ID: <3t8vi6msu8xx2vytorl4eyj3.1490302003801 at email.android.com>
>> Content-Type: text/plain; charset="utf-8"
>>
>> Dear Derek,I'll commit shortly a patch for meta-ggas. With the current
>> version of the code one cannot restart neither compute bands/dos with
>> mggas. If I manage I can try to add SCAN-libxc copying the routine for
>> tpss and tb09. By this token, it would be possible to implement many
>> other metaggas present in libxc, with little effort (help will be
>> welcome).
>> Best wishes.D.
>>
>>
>> Inviato da Tablet Samsung.
>> -------- Messaggio originale --------Da: "Vigil-Fowler, Derek"
>> <Derek.Vigil-Fowler at nrel.gov> Data: 23/03/17  21:23  (GMT+01:00) A:
>> q-e-developers at qe-forge.org Oggetto: Re: [Q-e-developers] SCAN functional
>> Hi Davide,
>> ????? Many thanks for letting me know that SCAN is in LibXC. I will try
>> that first. I am sorry to hear that meta-GGAs are not working well in QE.
>> I hope it does not cause problem for intended calculations. I?ll certainly
>> report any problems I have.
>>
>> -Derek
>>
>> On 3/23/17, 5:00 AM, "q-e-developers-bounces at qe-forge.org on behalf of
>> q-e-developers-request at qe-forge.org" <q-e-developers-bounces at qe-forge.org
>> on behalf of q-e-developers-request at qe-forge.org> wrote:
>>
>>> Re: SCAN functional
>>
>>
>> _______________________________________________
>> Q-e-developers mailing list
>> Q-e-developers at qe-forge.org
>> http://qe-forge.org/mailman/listinfo/q-e-developers
>> -------------- next part --------------
>> An HTML attachment was scrubbed...
>> URL:
>> http://qe-forge.org/pipermail/q-e-developers/attachments/20170323/55c74dc6
>> /attachment-0001.html
>>
>> ------------------------------
>>
>> Message: 5
>> Date: Thu, 23 Mar 2017 09:15:10 -0700
>> From: Mohammadreza Karamad <karamadr at gmail.com>
>> Subject: [Q-e-developers] Trouble with vc-relax: "Too many g vector"
>> 	error
>> To: q-e-developers at qe-forge.org
>> Message-ID:
>> 	<CABBh69ebGXtTo6ZYgt4wctU4wYtMDRXBdPsd3qeFXaXYhTtj1g at mail.gmail.com>
>> Content-Type: text/plain; charset="utf-8"
>>
>> Hi,
>>
>> I am attempting to optimize
>> ?a ?
>> unit cell for a H2MnO4 compound using vc-relax in ase-espresso. I have
>> got the following strange error:
>> "too many g-vectors"
>> I have used a e-cut of 45 Ry
>> ? which basically should be high enough?
>> .
>> ?I have also tried with different e-cut values, both smaller and
>> bigger. For the former ones, I get the error of "*g*-*vectors
>> missing"*?
>> ?. Although I am not sure,
>> I think this error has some thing to do with the e-cut.
>>
>> During cell relaxation, it seems that the size of the cell changes
>> significantly by ~10 Angstrom^3.
>> ? Moreover, it seems that the I got this error in the last step just
>> before performing scf calculations on the optimized structure, where
>> the calculations fail.?
>>
>> ? I have taken a cell_factor of 5,much larger than the default value,
>> i.e.,1.2. Below you can find the details of the calculators as well as
>> different parameters for vc-relax calculations. I would like to say
>> that I have used ase-espresso.
>>
>>    calculator = espresso(pw=600,
>>        dw=6000,
>>        xc='PBE',
>>        kpts = (6,6,6),
>>        nbands=-10,
>>        mode='vc-relax',
>>        cell_dynamics='bfgs',
>>        opt_algorithm='bfgs',
>>        cell_factor=5,
>>        spinpol=spinpol,
>>        outdir= outdir)
>>
>> Any help will be appreciated,
>>
>> Best,
>>
>> Reza
>> -------------- next part --------------
>> An HTML attachment was scrubbed...
>> URL:
>> http://qe-forge.org/pipermail/q-e-developers/attachments/20170323/96460a53
>> /attachment-0001.html
>>
>> ------------------------------
>>
>> _______________________________________________
>> Q-e-developers mailing list
>> Q-e-developers at qe-forge.org
>> http://qe-forge.org/mailman/listinfo/q-e-developers
>>
>>
>> End of Q-e-developers Digest, Vol 83, Issue 5
>> *********************************************
>
>
> _______________________________________________
> Q-e-developers mailing list
> Q-e-developers at qe-forge.org
> http://qe-forge.org/mailman/listinfo/q-e-developers
>

More information about the developers mailing list