[Pw_forum] Total energy diverges in cell relaxation
Hui Tang wrote: > Dear all, > > I'm trying to do cell relaxation on my Mg-Boron systems. For some of the > structures, I got extremely strange results. The total energy diverges > starting from some ionic step. For example, here are the total energy in > each ionic step for one of the problematic run: > Two suggestions 1) are atoms overlapping much more than they should ? I.e. are these new configurations reasonable, or are core overlaps very significant ? The pseudopotential approximation breaks down if cores overlap (a lot). 2) is it a bug/memory overrun ? Take the first configuration with much lower energy, and rerun that from scratch. Do you get an energy that looks like -41.7 Ry, or much lower ? nic - Prof Nicola Marzari Department of Materials Science and Engineering 13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu
[Pw_forum] Total energy diverges in cell relaxation
Hi Axel, I have used a much larger cutoff energy to rerun a problematic calculation. This rerun does not have the divergence problem any more. I'm going to do a few more runs to confirm this. I think it's just as what you said that the effective cutoff energy becomes smaller as the cell expands. So I probably should check the convergence of stresses in my largest cell with respect to the number of plane waves. btw, I did ghost states tests for my pseudo-potentials and there isn't any. p.s., I remember to put on my affiliation this time. :) Thank you very much, - Hui Tang Graduate Student Department of Applied Physics Yale University Email: hui.tang at yale.edu Phone: 1-203-432-7763 On Thu, Oct 16, 2008 at 4:00 PM, Axel Kohlmeyer wrote: > On Thu, 16 Oct 2008, Hui Tang wrote: > > hi hui, > > your cell is expanding quite a bit and your cutoff is not > that large. with an expanding cell, your _effective_ cutoff > becomes much smaller as the number of plane waves remains > constant during a variable cell calculation. have you tested > the stess tensor convergence vs. the cutoff? you basically > have to figure out how many plane waves you need for a well > converged stress tensor at the largest expansion of your cell > and then you have to set the cutoff at the beginning so that > it results in the same number of plane waves... > > that would explain why it works (better) if you restart. > > also, did you test the pseudopotentials for ghost states? > > cheers, > axel. > > p.s.: please provide an affiliation. thanks. > > HT> Hi Nic, > HT> > HT> 1) I have tested the pseudo-potentials (psps) with small atomic > HT> separations. The pseudo-potentials give reasonable till two atoms are > as > HT> close as 0.8 A. When the total energy starts to diverge, which is about > HT> Ionic iteration 17, the smallest inter-atomic distance is about 1.4 A. > So I > HT> do not think the breakdown of the psps is the trigger of this > divergence of > HT> the total energy, although it should be the reason for the enormous > large > HT> negative energies. > HT> > HT> 2) I have tried to restart the calculation from where the divergence > starts. > HT> Once I did that, the first scf calculation would give a reasonable > energy > HT> around -41.7 Ry, and the calculation could run smoothly till the total > HT> energy converges most of the times. So what I deal with this problem > right > HT> now is to restart calculations from where they start to have diverged > total > HT> energies and it's been working well. However, I hope this is not the > optimal > HT> way to solve this problem. :) > HT> > HT> Thank you very much, > HT> Hui > HT> > > -- > === > Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu > Center for Molecular Modeling -- University of Pennsylvania > Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323 > tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425 > === > If you make something idiot-proof, the universe creates a better idiot. > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20081016/b6730b73/attachment.htm
[Pw_forum] possible bug in scatter_forw.f90
Dear Alexandar,
Thank you very much. Your answers make sense and now i have better
understanding of the subroutine scatter_forw.f90. I have one question
about poten.f90 subroutine, where you calculate the V(p,gper) in each
slices of the slab(equation 15). I have a question about Gz, which is
2*pi*q/L, with [-N/2]+1<=q>=[N/2]. Now in the code at the line at the line
60, you have Gz multiplied with bg(3,3). I kind of understand this is
because 3rd direction is perpendicular to xy plane, but not totally clear
about it. Would you mind explaining? my other question is when reciprocal
space vectors b1, b2 and b3 are not orhogonal to each other, which is
more general case, then should not we take bg(1,3) and bg(2,3) in there
too?
Thank you once again for listening to my questions and i hope i am
not bothering you too much.
Regards,
Manoj Srivastava
Physics Graduate Student
Department of Physics
Gainesville, FL, USA.
have On Wed, 15 Oct 2008, Alexander wrote:
> Dear Manoj
>
> On Tuesday 14 October 2008 17:49, Manoj Srivastava wrote:
> > Dear Alexader,
> >
> > Thank you very much for your reply. I was out of town and could not see
> > the message till yesterday. I have some questions on it, it would be great
> > if you could answer them.
> >
> > On Fri, 10 Oct 2008, Alexander wrote:
> > > Dear Manoj
> > >
> > > On Friday 10 October 2008 05:20, Manoj Srivastava wrote:
> > > > Any follow ups on the message below? Alexander?
> > > >
> > > > Regards,
> > > > Manoj
> > > >
> > > > On Wed, 8 Oct 2008, Manoj Srivastava wrote:
> > > > > Dear All,
> > > > > Is any body familiar with the scatter_forw.f90 subroutine of PWCOND?
> > > > > I think there is a bug in this subroutine at the place where you
> > > > > calculate intw2, which is z integration of nonlocal wavefunction with
> > > > > beta function. I have looked up the necessary formulae in the Choi &
> > > > > Ihm's paper (PRB 59, 2267, Jan 1999). In the paper, nonlocal
> > > > > wavefunction has 3 terms, one of which contains beta function, say W.
> > > > > The other two parts dont have any W in them, rather they are just
> > > > > plane wave solution. (Please have a look at equation 24 and 26 of the
> > > > > paper ). So, when we are doing z integration of nonlocal wavefunction
> > > > > with beta function W, we should have three terms, one of which should
> > > > > contain two W, but rest should just have one W. On the other hand in
> > > > > the code all three terms contain two beta functions!! (please have a
> > > > > look at line 228 of scatter_forw.f90). I am wondering if my
> > > > > understanding is right?
> > >
> > > Your understanding is wrong. On the line 228 you add to the integral ONLY
> > > contribution with two beta functions (from the 1st term in the nonlocal
> > > function, see Eq. 24) but on the line 393 the contribution from the 2nd
> > > term is added too. There is no 3rd term since every step you rotate your
> > > local solutions \psi_n in such a way that b_{\lambda \alpha lm} vanish
> > > (see the paragraph after Eq. 37).
> >
> > I believe that the equations 40, 41 and 42 implemented in the code are
> > in the momentum space. In the code, intw2 has 3 terms. We can tell this by
> > looking at the subroutine integrals.f90, where int2d is defined. Now, by
> > looking at the choi& Ihm 's paper, the very first term of psi{alpha,l,m}
> > has f{lambda,alpha,l,m}(equation26).And, if we substitute this in equation
> > 40, we should just get one term. I am just confuesd on the fact that how
> > come this one term in paper gets transformed into 3 terms of the code!
>
> intw2(alpha,beta) is the integral of nonlocal solution (24) with w functions,
> intw2(alpha,beta) = \int_{0,d} w_{alpha}(r)*psi_{beta}(r)dr, where
> alpha,beta={alpha,l,m} in the paper. In each slab it will have TWO
> contributions from the first and second terms of (24), the one from the third
> term vanishes as I explained before. These two contributions are calculated
> on the lines 228 and 393 of the code. The integrals intw1
> (alpha,n)=\int_{0,d} w_{alpha}(r)*psi_{n}(r)dr are the integrals of local
> solutions (17) with the same projector functions w.
> The integrals intw2 and intw1 have nothing to do with boundary conditions
> (40). They enter in (11) and are needed to construct the final solutions
> according to Eq.(10). Substituting the form of this solution (10) in the
> boundary conditions (40) you will be led to the generalized eigenvalue
> problem, AX=exp(ikd) BX, with known matrices A and B and unknown
> coefficients X={a_n,C_{alpha,l,m}} defining the resulting solution (10).
>
> > Would you mind explaining? Also, the integrals that are defined in the
> > code are only in the slabs (nz1), so how are you taking the integration in
> > the region (0,d), which we need according to equations 40, 41 and 42 ?
>
> When you go over slabs you add new contributions to the integrals intw1 and
> intw2 (see the line 228 for the first contribution to intw2 and
[Pw_forum] DOS plot
Dear Sang-Hwan: In addition to Lan Haiping's words, I introduce you a easy method to plot DOS with Ef=0eV: (1) Find the Ef in the output file of self consistent calculation(e.g. Ef=0.0345Ry); (2) plot DOS with gnuplot : plot '*.DOS' using ($1-0.0345):2 with line where the 1st column should be Energy and the 2nd column should be the corresponding DOS. with ($1-0.0345), the all energy values will go down for 0.0345Ry together, letting the fermi level coincide with 0 Ry in the graph. Best Wishes! Yours Sincerely L.F.Huang -- == L.F.Huang(???) lfhuang at theory.issp.ac.cn == Add: Research Laboratory for Computational Materials Sciences, Instutue of Solid State Physics,the Chinese Academy of Sciences, P.O.Box 1129, Hefei 230031, P.R.China Tel: 86-551-5591464-328(office) Fax: 86-551-5591434 Web: http://theory.issp.ac.cn (website of our theory group) http://www.issp.ac.cn(website of our institute) == > Date: Wed, 15 Oct 2008 23:02:14 -0500 > From: ski2 at mail.uh.edu > Subject: [Pw_forum] DOS plot > To: pw_forum at pwscf.org > Message-ID: > Content-Type: text/plain; charset="us-ascii" > > Dear All, > > I have two simple questions about Fermi level and DOS. > For insulator electronic structure calculations, is the Fermi level evaluated > set to the top of a valence band? > When DOS plotted, is the Fermi level manually set to zero by shifting the > value achieved self consistently in a scf.out file? > > Best Wishes, > Sang-Hwan > > > Department of Chemistry > University of Houston > 136 Fleming Building > Houston, TX 77204-5003 > -- next part -- > An HTML attachment was scrubbed... > URL: > http://www.democritos.it/pipermail/pw_forum/attachments/20081015/b96348a6/attachment.html > > > -- > > Message: 2 > Date: Thu, 16 Oct 2008 14:05:22 +0800 > From: "lan haiping" > Subject: Re: [Pw_forum] DOS plot > To: "PWSCF Forum" > Message-ID: > > Content-Type: text/plain; charset="iso-8859-1" > > Hi, > Actually, you should align with fermi level yourself when preparing DOS > plot. > > Regards > > > On Thu, Oct 16, 2008 at 12:02 PM, wrote: > > > Dear All, > > > > I have two simple questions about Fermi level and DOS. > > For insulator electronic structure calculations, is the Fermi level > > evaluated set to the top of a valence band? > > When DOS plotted, is the Fermi level manually set to zero by shifting the > > value achieved self consistently in a scf.out file? > > > > Best Wishes, > > Sang-Hwan > > > > > > Department of Chemistry > > University of Houston > > 136 Fleming Building > > Houston, TX 77204-5003 > > ___ > > Pw_forum mailing list > > Pw_forum at pwscf.org > > http://www.democritos.it/mailman/listinfo/pw_forum > > > > > > > -- > Hai-Ping Lan > Department of Electronics , > Peking University , Bejing, 100871 > lanhaiping at gmail.com, hplan at pku.edu.cn > -- next part -- > An HTML attachment was scrubbed... > URL: > http://www.democritos.it/pipermail/pw_forum/attachments/20081016/8b57ac6f/attachment-0001.htm > > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20081016/9f154731/attachment.htm
[Pw_forum] post doc opening in Padova
A POSTDOCTORAL POSITION of 2 YEARS (with the possibility of renewal) is available from February/March 2009 at the Group of Theoretical Condensed-Matter Physics, Department of Physics "Galileo Galilei" of the University of Padova, Italy. APPLICATION DEADLINE: Dic 08, 2008 RESPONSIBILITIES: To investigate with ab initio computational methods surface processes, such as physisorption of rare gas atom on metal surface, molecule adsorption, and interaction of water with hydrophobic/hdrophilic surfaces. Ab initio methods will be based on Density Functional Theory with inclusion of Wan der Waals effects (see Phys. Rev. Lett., 100, 053002 (2008)). The research group, which is part of the DEMOCRITOS National Simulation Center (Trieste, Italy), is currently composed of three professors of the University of Padova (Flavio Toigo, Francesco Ancilotto, and Pier Luigi Silvestrelli). Applications should be accompanied by a detailed CV and a publication list (recommendation letters are appreciated). After a preliminary, informal selection, suitable candidate(s) will be invited to submit a formal application to the University of Padova in order to be evaluated with a colloquium/seminar at our Institute, in January 2009. N.B. Travel expenses cannot be reimbursed. REQUIREMENTS: Dynamic and motivated scientist with Ph.D. in theoretical physics, chemistry or related fields. Candidates with a documented experience in scientific programming, high-performance computing, and application of ab initio packages, such as Quantum-ESPRESSO(PWscf) and CPMD, are strongly preferred. SALARY: about 1450 Euro per month (net). CONTACT: Prof. Pier Luigi Silvestrelli E-mail: psil at pd.infn.it Phone: +39 049 8277171 Fax: +39 049 8277102 Address: Department of Physics "Galileo Galilei", University of Padova, Via Marzolo 8, 35131 Padova, Italy
[Pw_forum] Total energy diverges in cell relaxation
On Thu, 16 Oct 2008, Hui Tang wrote: hi hui, your cell is expanding quite a bit and your cutoff is not that large. with an expanding cell, your _effective_ cutoff becomes much smaller as the number of plane waves remains constant during a variable cell calculation. have you tested the stess tensor convergence vs. the cutoff? you basically have to figure out how many plane waves you need for a well converged stress tensor at the largest expansion of your cell and then you have to set the cutoff at the beginning so that it results in the same number of plane waves... that would explain why it works (better) if you restart. also, did you test the pseudopotentials for ghost states? cheers, axel. p.s.: please provide an affiliation. thanks. HT> Hi Nic, HT> HT> 1) I have tested the pseudo-potentials (psps) with small atomic HT> separations. The pseudo-potentials give reasonable till two atoms are as HT> close as 0.8 A. When the total energy starts to diverge, which is about HT> Ionic iteration 17, the smallest inter-atomic distance is about 1.4 A. So I HT> do not think the breakdown of the psps is the trigger of this divergence of HT> the total energy, although it should be the reason for the enormous large HT> negative energies. HT> HT> 2) I have tried to restart the calculation from where the divergence starts. HT> Once I did that, the first scf calculation would give a reasonable energy HT> around -41.7 Ry, and the calculation could run smoothly till the total HT> energy converges most of the times. So what I deal with this problem right HT> now is to restart calculations from where they start to have diverged total HT> energies and it's been working well. However, I hope this is not the optimal HT> way to solve this problem. :) HT> HT> Thank you very much, HT> Hui HT> -- === Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu http://www.cmm.upenn.edu Center for Molecular Modeling -- University of Pennsylvania Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323 tel: 1-215-898-1582, fax: 1-215-573-6233, office-tel: 1-215-898-5425 === If you make something idiot-proof, the universe creates a better idiot.
[Pw_forum] Total energy diverges in cell relaxation
Hi Nic, 1) I have tested the pseudo-potentials (psps) with small atomic separations. The pseudo-potentials give reasonable till two atoms are as close as 0.8 A. When the total energy starts to diverge, which is about Ionic iteration 17, the smallest inter-atomic distance is about 1.4 A. So I do not think the breakdown of the psps is the trigger of this divergence of the total energy, although it should be the reason for the enormous large negative energies. 2) I have tried to restart the calculation from where the divergence starts. Once I did that, the first scf calculation would give a reasonable energy around -41.7 Ry, and the calculation could run smoothly till the total energy converges most of the times. So what I deal with this problem right now is to restart calculations from where they start to have diverged total energies and it's been working well. However, I hope this is not the optimal way to solve this problem. :) Thank you very much, Hui On Thu, Oct 16, 2008 at 3:09 PM, Nicola Marzari wrote: > Hui Tang wrote: > > Dear all, > > > > I'm trying to do cell relaxation on my Mg-Boron systems. For some of the > > structures, I got extremely strange results. The total energy diverges > > starting from some ionic step. For example, here are the total energy in > > each ionic step for one of the problematic run: > > > > Two suggestions > > 1) are atoms overlapping much more than they should ? I.e. are these > new configurations reasonable, or are core overlaps very significant ? > The pseudopotential approximation breaks down if cores overlap (a lot). > > 2) is it a bug/memory overrun ? Take the first configuration with > much lower energy, and rerun that from scratch. Do you get an > energy that looks like -41.7 Ry, or much lower ? > >nic > > > > - > Prof Nicola Marzari Department of Materials Science and Engineering > 13-5066 MIT 77 Massachusetts Avenue Cambridge MA 02139-4307 USA > tel 617.4522758 fax 2586534 marzari at mit.edu http://quasiamore.mit.edu > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- next part ------ An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20081016/b17a8b89/attachment.htm
[Pw_forum] Total energy diverges in cell relaxation
correction = 0.000207 Total force = 1027.278240 Total SCF correction = 0.000338 For this problematic system, this behavior is very stable. No matter I use 'cg' or 'david', change the mixing parameter from 0.9 to 0.1, I still get similar problem. I cannot find anything relevant in the archive. I wonder if anyone knows what's going on here and can help me get rid of this annoying problem. btw, the input, output files and pseudo-potentials can be found at: http://volga.eng.yale.edu/hui/MgonBS.tar.gz Thank you very much, - Hui Tang Graduate Student Department of Applied Physics Yale University Email: hui.tang at yale.edu Phone: 1-203-432-7763 -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20081016/84029706/attachment.htm
[Pw_forum] DOS plot
Hi, Actually, you should align with fermi level yourself when preparing DOS plot. Regards On Thu, Oct 16, 2008 at 12:02 PM, wrote: > Dear All, > > I have two simple questions about Fermi level and DOS. > For insulator electronic structure calculations, is the Fermi level > evaluated set to the top of a valence band? > When DOS plotted, is the Fermi level manually set to zero by shifting the > value achieved self consistently in a scf.out file? > > Best Wishes, > Sang-Hwan > > > Department of Chemistry > University of Houston > 136 Fleming Building > Houston, TX 77204-5003 > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > > -- Hai-Ping Lan Department of Electronics , Peking University , Bejing, 100871 lanhaiping at gmail.com, hplan at pku.edu.cn -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20081016/8b57ac6f/attachment.htm
[Pw_forum] Antiferro
Dear Naol, Your problem is probably what it says in the error, that is you have overlapping atoms. You have either put in wrong atomic positions or are confused about how to use parameters and options such as celldm, alat etc. In either case I suggest using xcrysden to visualize your structure before submitting it. Atomic positions should not be guessed but should be taken from some preexisting data, try someting like : http://cst-www.nrl.navy.mil/lattice/ Best, Hande On Thu, 16 Oct 2008, Naol Regassa wrote: > Dear sir we were doing our scf calculation on peroviskites (CaMnO3) > and we know that it is fcc or orthorohmbic G-type antiferro with 10 > atoms. When we prepare our input try to excute it tells us that we > have an error like 'wrong atomic positions(coordinates)' or 'existance > of overlaping of atoms'. What should we do to find(guess) the atomic > positions? > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > -- Hande Toffoli Department of Physics Office 439 Middle East Technical University Ankara 06531, Turkey Tel : +90 312 210 3264 http://www.physics.metu.edu.tr/~hande
[Pw_forum] Antiferro
Dear sir we were doing our scf calculation on peroviskites (CaMnO3) and we know that it is fcc or orthorohmbic G-type antiferro with 10 atoms. When we prepare our input try to excute it tells us that we have an error like 'wrong atomic positions(coordinates)' or 'existance of overlaping of atoms'. What should we do to find(guess) the atomic positions?
[Pw_forum] Antiferro
Dear sir we were doing our scf calculation on peroviskites (CaMnO3) and we know that it is fcc or orthorohmbic G-type antiferro with 10 atoms. When we prepare our input try to excute it tells us that we have an error like 'wrong atomic positions(coordinates)' or 'existance of overlaping of atoms'. What should we do to find(guess) the atomic positions.
[Pw_forum] Bug in axsf generation?
Thank you for taking the time. My cp.x file was bad to begin with. Sorry to pollute the list like that. Jan Beck Txl Group > Date: Tue, 14 Oct 2008 10:45:24 +0200 > From: Tone Kokalj > Subject: Re: [Pw_forum] Bug in axsf generation? > To: PWSCF Forum > Message-ID: <1223973924.1455.9.camel at walk.ijs.si> > Content-Type: text/plain > > On Fri, 2008-10-10 at 08:23 -0600, Jan Beck wrote: >> I have an axsf file that I have generated with quantum espresso's cppp. I >> have checked with the specification of the axsf file format, and the >> output from quantum espresso is consistent with that. > >> WARNING: Atom 2 and atom 6 overlap !!! >> Atom 6 deleted !!! >> WARNING: Atom 2 and atom 10 overlap !!! >> Atom 10 deleted !!! > > I have looked at your file. It is nothing wrong with its syntax, but the > structure is a complete mass: there are many overlapping atoms (look > into the file). Therefore you receive the atom-deleted warnings! > > Either your original cp.x file already contains massy structure, or the > cppp utility has a bug (i never used the latter). > > Regards, Tone > -- > Tone Kokalj > J. Stefan Institute, Jamova 39, 1000 Ljubljana, Slovenia (tel: > +386-1-477-3523 // fax:+386-1-477-3822) > >
[Pw_forum] DOS plot
Dear All, I have two simple questions about Fermi level and DOS. For insulator electronic structure calculations, is the Fermi level evaluated set to the top of a valence band? When DOS plotted, is the Fermi level manually set to zero by shifting the value achieved self consistently in a scf.out file? Best Wishes, Sang-Hwan Department of Chemistry University of Houston 136 Fleming Building Houston, TX 77204-5003 -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20081015/b96348a6/attachment.htm
