[Pw_forum] Four Electron Pseudopotnetial for tin: US vs NC-PP
On Tue, 08 May 2012 19:45:38 +0200, Giovani Faccin wrote: > 1 - On the 6D line, why a negative number of electrons? Sorry if I'm > mistaken, but shouldn't that number be positive? No, this is correct, it only means that the state label is reserved but not used in the all-electron calculation. However, I would rather use the 5D instead of the 6D, but I does not really make any difference for the pseudo-wavefunctions. > 2 - On this same line, why not use the reference energy (0.00), since > it's a bound state, as done in 5S and 5P? Because using a negative occupation number the eigenvalue is not computed :-) > > 3 - If everything else is ok, the cutoff parameters (the 2.50 and 2.70 > guys in the 6th and 7th columns) are usually to blame when generation > fails. Indeed, I think it is almost impossible to find a good cutoff radius for all the 5 reference wavefunctions, maybe one could start with a single-channel pseudopotential; then only add a second channel if necessary. However, 2 channels are required for ultrasoft, maybe Sven really want to make an USPP in the end. bests -- Lorenzo Paulatto IdR CNRS @ IMPMC/UPMC - Paris 6 phone: +33 (0)1 44275 084 / skype: paulatz www: http://www.impmc.upmc.fr/~paulatto/ mail: 23-24/4?14 Bo?te courrier 115, 4 place Jussieu 75252 Paris C?dex 05
[Pw_forum] cutoff convergence of sapphire for elastic properties
On 08/05/2012 18:42, J?rg Buchwald wrote: > But to get the elastic constants in the elastic regime, i would > like then apply strains of serveral per mill, which is of the same > order of magnitude, i.e. also corresponds to changes of the stress > tensor of the order of 1 kbar, which means that errors in the kbar > range would be too high. Hi Joerg, relative quantities (in this case, difference in stress upon change in strain) typically converge much faster than absolute quantities (the stress for a given strain), so you should be fine. Worth checking. > An alternative could be the calculation of the elastic constants using > the second derivative of the energy. But this won't work for big > supercells due to the computation time and the number of measuring > points needed. I think they would be comparable - for one elastic constant using the stress tensor I would probably do three calculations - one around equilibrium, and plus or minus 0.5% that . Using the energy, maybe five? +1 +0.5 0 -.5 -1? nicola -- Prof Nicola Marzari, Chair of Theory and Simulation of Materials, EPFL
[Pw_forum] cutoff convergence of sapphire for elastic properties
Am Mon, 7 May 2012 14:22:58 -0400 schrieb Mike Mehl : > To follow up on Nicola's point, the bulk modulus of sapphire is 240 > GPa or 2400 kbar > (http://www.mt-berlin.com/frames_cryst/descriptions/sapphire.htm). So > a 1 kbar error corresponds to a very small change in volume. > > If we use the quick and dirty Birch equation of state: > > P(V) = 3/2 K0 [ (V0/V)^(7/3) - (V0/V)^(5/3)] > > with K0 = 2400 kbar and ask what volume will produce a 1 kbar change > in pressure we get > > delta V/V0 = +/- 0.0004 > > Considering the normal errors in DFT, it's not worth trying to > converge the stress to the 1 kbar accuracy you're trying to achieve. But to get the elastic constants in the elastic regime, i would like then apply strains of serveral per mill, which is of the same order of magnitude, i.e. also corresponds to changes of the stress tensor of the order of 1 kbar, which means that errors in the kbar range would be too high. An alternative could be the calculation of the elastic constants using the second derivative of the energy. But this won't work for big supercells due to the computation time and the number of measuring points needed. > > On Mon, May 7, 2012 at 10:07 AM, Nicola > Marzari wrote: > > > > > How much does 1 kbar error translates into an error in lattice > > parameter? (keep atoms fixed, using relative coordinates, cutoff > > fixed, and expand celldm(1) by 0.3% - what's the change in stress? > > that change should be very well converged) > > > > -- > Michael Mehl > US Naval Research Laboratory > Washington DC > (Home email)
[Pw_forum] regarding weight factor
On Fri, 2012-05-04 at 21:04 +0530, himanshu at iopb.res.in wrote: > I am confused about the weight factor. the weight factor for manually supplied k-point grids is determined by the symmetry: it is proportional to the number of vectors in the star of each k-point P. -- Paolo Giannozzi, IOM-Democritos and University of Udine, Italy
[Pw_forum] HSE problem
On Mon, 2012-05-07 at 11:21 +0100, debbichi mourad wrote: > Why such a problem occurs? which problem? absolute values of the Fermi energy are arbitrary. P. -- Paolo Giannozzi, IOM-Democritos and University of Udine, Italy
[Pw_forum] Four Electron Pseudopotnetial for tin: US vs NC-PP
Thank you Lorenzo! I realized what I'd written made no sense right after sending the email. Was typing another one to correct that, but you were faster. It's the after lunch sleepness :O) Giovani 2012/5/8 Lorenzo Paulatto > On Tue, 08 May 2012 19:45:38 +0200, Giovani Faccin > wrote: > > 1 - On the 6D line, why a negative number of electrons? Sorry if I'm > > mistaken, but shouldn't that number be positive? > > No, this is correct, it only means that the state label is reserved but > not used in the all-electron calculation. However, I would rather use the > 5D instead of the 6D, but I does not really make any difference for the > pseudo-wavefunctions. > > > 2 - On this same line, why not use the reference energy (0.00), since > > it's a bound state, as done in 5S and 5P? > > Because using a negative occupation number the eigenvalue is not computed > :-) > > > > > 3 - If everything else is ok, the cutoff parameters (the 2.50 and 2.70 > > guys in the 6th and 7th columns) are usually to blame when generation > > fails. > > Indeed, I think it is almost impossible to find a good cutoff radius for > all the 5 reference wavefunctions, maybe one could start with a > single-channel pseudopotential; then only add a second channel if > necessary. However, 2 channels are required for ultrasoft, maybe Sven > really want to make an USPP in the end. > > bests > > -- > Lorenzo Paulatto IdR CNRS @ IMPMC/UPMC - Paris 6 > phone: +33 (0)1 44275 084 / skype: paulatz > www: http://www.impmc.upmc.fr/~paulatto/ > mail: 23-24/4?14 Bo?te courrier 115, 4 place Jussieu 75252 Paris C?dex 05 > ___ > 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/20120508/a87ad1de/attachment.htm
[Pw_forum] Four Electron Pseudopotnetial for tin: US vs NC-PP
Dear Sven, I've built a few pseudopotentials in the past to study hybrid clusters. On those calculations I found that the TM pseudization succeeds at generating pseudopotentials much more often than the rrkj pseudization. Perhaps you could try it, and if it works, then use the cutoff parameters data that worked out with TM as a guide to try to generate a pseudo using rrkj, should you really have to stick to it. That said, a few ideas that might help: 1 - On the 6D line, why a negative number of electrons? Sorry if I'm mistaken, but shouldn't that number be positive? 2 - On this same line, why not use the reference energy (0.00), since it's a bound state, as done in 5S and 5P? 3 - If everything else is ok, the cutoff parameters (the 2.50 and 2.70 guys in the 6th and 7th columns) are usually to blame when generation fails. You'll have to explore those parameters in order to find a good combination. To aid in that, plot the all electron wavefunctions generated and look at them. That will give you an idea of reasonable values to use for the cutoff. Best, Giovani M. Faccin UFMS/Brazil 2012/5/8 Sven Heiles > Dear Quantum Espresso users. > > Our group is dealing with the global optimizations of clusters using > Quantum Espresso. In the moment we investigate > the Mn-Sn system for which it is known that a considerable charge > transfer between Mn(positive) and Sn(negative) occurs. In order to model > the clusters in a finite time it would be useful only to use a PP for > the four valence electrons of tin while for Mn it is no problem > to use 15 electrons (only one or two Mn are used). The Mn-PP is > generated using the rrkjus fitting routine but the Sn-PP generation > always fails if only four electrons > are used for the ultra-soft PP. We tested to generate the 14 electron > US-PP for tin in analogy to the Pb-US-PP found in the QE PP library and > were able to generate a PP. > Unfortunately, that many electrons are not of much use in our global > optimization calculations. > On the other hand the generation of a NC-PP with four electrons for tin > is able to reproduce bulk and > dimer properties but a similar procedure for Mn is not of much use. Does > someone has an idea what the problem for the 4 electron > US-PP for tin could be? Please find the generation parameters below. > > All the best > > Sven > > > title='Sn', > zed=50.0, > rel=1, > iswitch=3, > config='[Kr] 4d10.0 5s2.0 5p2.0 6d-2.0', > dft='LDA', > / > > lloc=2, > pseudotype=3, > file_pseudopw='Sn.pz-rrkjus.UPF', > author='', > / > 5 > 5S 1 0 2.00 0.00 2.50 2.50 > 5S 1 0 0.00 0.05 2.50 2.50 > 5P 2 1 2.00 0.00 2.70 2.90 > 5P 2 1 0.00 0.05 2.70 2.90 > 6D 3 2 -2.00 0.15 2.70 2.70 > > > -- > Dipl. Ing. Sven Heiles > Technische Universit?t Darmstadt > AK Sch?fer > Eduard-Zintl-Institut > Petersenstra?e 20 > D-64287 Darmstadt ? Germany > > Phone: ++49-(0)6151-164397 > Fax: ++49-(0)6151-166024 > Web:http://www.tu-darmstadt.de/fb/ch/cluster/schaefer.tud > > ___ > 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/20120508/787b8748/attachment-0001.htm
[Pw_forum] cutoff convergence of sapphire for elastic properties
On Tue, May 8, 2012 at 12:42 PM, J?rg Buchwald wrote: >> Considering the normal errors in DFT, it's not worth trying to >> converge the stress to the 1 kbar accuracy you're trying to achieve. > > But to get the elastic constants in the elastic regime, i would > like then apply strains of serveral per mill, which is of the same > order of magnitude, i.e. also corresponds to changes of the stress > tensor of the order of 1 kbar, which means that errors in the kbar > range would be too high. question is: do you have the same absolute error of the same magnitude and sign on all stress tensor computations, so that the *relative* error would be much smaller? remember, that when you do e.g. a geometry optimization you use forces (derivatives of the potential energy) that are converged long before your energy is converged, so by your argument, you would have to use a much higher cutoff for that as well. > An alternative could be the calculation of the elastic constants using > the second derivative of the energy. But this won't work for big > supercells due to the computation time and the number of measuring > points needed. see the comment about slow and fast converging properties. you probably don't need to converge the energy as tightly for this due to error cancellation, too. however, convergence and error cancellation is not always easily predictable. some careful experimentation is often better than using either brute force, guessing or faith in some random person's ramblings. ;-) cheers, axel. > > >> >> On Mon, May 7, 2012 at 10:07 AM, Nicola >> Marzari wrote: >> >> > >> > How much does 1 kbar error translates into an error in lattice >> > parameter? (keep atoms fixed, using relative coordinates, cutoff >> > fixed, and expand celldm(1) by 0.3% - what's the change in stress? >> > that change should be very well converged) >> > >> > -- >> Michael Mehl >> US Naval Research Laboratory >> Washington DC >> (Home email) > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum -- Dr. Axel Kohlmeyer akohlmey at gmail.com ?http://goo.gl/1wk0 College of Science and Technology Temple University, Philadelphia PA, USA.
[Pw_forum] Four Electron Pseudopotnetial for tin: US vs NC-PP
On Tue, 08 May 2012 11:43:22 +0200, Sven Heiles wrote: > Does someone has an idea what the problem for the 4 electron > US-PP for tin could be? Please find the generation parameters below. Dear Sven, first of all have a look at the ps library http://qe-forge.org/frs/?group_id=35_id=109#pslibrary-pslibrary-0-2-4-title-content database by Dal Corso, it may already contain what you are looking for. If it does not, I have to admin that I got a bit lost in your message. What exactly do you want to do? An ultrasoft-pseudopotential for Sn with 4 valence electrons? bests -- Lorenzo Paulatto IdR CNRS @ IMPMC/UPMC - Paris 6 phone: +33 (0)1 44275 084 / skype: paulatz www: http://www.impmc.upmc.fr/~paulatto/ mail: 23-24/4?14 Bo?te courrier 115, 4 place Jussieu 75252 Paris C?dex 05
[Pw_forum] too many bands are not converged
Dear all When I run "Relax" , after 12 iteration run stoped and show me : "from c_bands : error # 1 too many bands are not converged stopping ..." thanks a lot Bani Adam Faculty of Science Egypt
[Pw_forum] Regarding anharmonic IFCs
On Tue, 08 May 2012 08:07:56 +0200, david yang wrote: > (1)--Whether my understanding is correct? Cross check with my post of February: http://www.democritos.it/pipermail/pw_forum/2012-February/023255.html > (2)--For my work i need anharmonic IFCs in real space means phi(0,l,l') > I known that this can be done by taking inverse Fourier transform of > phi(0,q',-q') , but i am clue less after this . > Can anyone give me hint that how to proceed for inverse Fourier transform Like in the q2r.x code (PH/q2r.f90) but with one more index bests -- Lorenzo Paulatto IdR CNRS @ IMPMC/UPMC - Paris 6 phone: +33 (0)1 44275 084 / skype: paulatz www: http://www.impmc.upmc.fr/~paulatto/ mail: 23-24/4?14 Bo?te courrier 115, 4 place Jussieu 75252 Paris C?dex 05
[Pw_forum] Regarding anharmonic IFCs
Dear Users and developers of Quantum espresso, I have some question regarding D3 code . I am interested in Anharmonic properties of crystals and for that i am using D3 code of quantum espresso while going through the source code i figured out that present version of quantum espresso calculates Anharmonic IFCs only for q=0,q' and q" . So that mean that quantum espresso gives us anharmonic IFCs in reciprocal space something like phi(0,q',q') where phi denotes anharmonic IFC. I have two questions on this (1)--Whether my understanding is correct? (2)--For my work i need anharmonic IFCs in real space means phi(0,l,l') I known that this can be done by taking inverse Fourier transform of phi(0,q',-q') , but i am clue less after this . Can anyone give me hint that how to proceed for inverse Fourier transform ... I will be great-full to all of you. Awaiting for responses Thanks a lot With warm regards David yang NTU Singapore -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20120508/b6846ead/attachment.htm