Re: [Pw_forum] Converged SCF total energies different in 6.0/6.1/6.2
Thanks very much for this important information. Best regards, Feng >See this note in file Doc/release-notes: > > * Some constants in the definition of PBE functionals were truncated to >6 significant digits. While not a bug, this could lead to tiny differences >with respect to previous results and other XC implementations (r13592) > >More explicitly: some constants that previously were truncated to 6 digits, are now defined with full precision. This removes minor numerical discrepancies with the results obtained using libxc. > >Paolo >On Thu, Dec 21, 2017 at 3:11 AM, WF <kalamaill...@gmail.com> wrote: >> Dear all, >> >> In tests a normal DFT calculations with Si, I noticed that QE 6.0 and >> 6.1 gives exactly same total energy with 6.0 and 6.1 (-15.75190880 >> Ry), but with QE 6.2.1, the total energy is -15.75191309Ry. I tested >> this on two different machines with different compilers, the results >> are the same across all of them (at least to last digit). I did some >> tests on larger systems also and found even larger difference. This >> 1e-5 Ry order of difference is not negligible. Is there anything >> introduced in 6.2 leads to this difference? Thanks very much. >> >> >> >> Best regards, >> >> Feng >> >> >> >> >> ___ >> Pw_forum mailing list >> Pw_forum@pwscf.org >> http://pwscf.org/mailman/listinfo/pw_forum >> > > > >-- >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 > -- next part -- > An HTML attachment was scrubbed... > URL: http://pwscf.org/pipermail/pw_forum/attachments/20171221/7c532d75/attachment -0001.html -- Dr. Feng Wu Postdoctoral researcher Department of Chemistry and Biochemistry University of California, Santa Cruz Phone: 831-459-2874 ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
[Pw_forum] Converged SCF total energies different in 6.0/6.1/6.2
Dear all, In tests a normal DFT calculations with Si, I noticed that QE 6.0 and 6.1 gives exactly same total energy with 6.0 and 6.1 (-15.75190880 Ry), but with QE 6.2.1, the total energy is -15.75191309Ry. I tested this on two different machines with different compilers, the results are the same across all of them (at least to last digit). I did some tests on larger systems also and found even larger difference. This 1e-5 Ry order of difference is not negligible. Is there anything introduced in 6.2 leads to this difference? Thanks very much. Best regards, Feng The input file is attached: calculation='scf' outdir='./tmp' prefix='si' wf_collect = .true., / ibrav=2, celldm(1) = 10.263, nat=2, ntyp= 1, ecutwfc = 40, / mixing_mode='plain' mixing_beta = 0.7, conv_thr = 1.0d-10, / ATOMIC_SPECIES Si 1.0 Si_ONCV_PBE-1.1.upf ATOMIC_POSITIONS {alat} Si 0.00 0.00 0.00 Si 0.25 0.25 0.25 K_POINTS (automatic) 4 4 4 0 0 0 == -- Dr. Feng Wu Postdoctoral researcher Department of Chemistry and Biochemistry University of California, Santa Cruz Phone: 831-459-2874 ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
Re: [Pw_forum] Band parallelization of EXX ACE in QE 6.1
Thanks, this is a small case (16 Silicon) with ONCV pseudo potential. Run with 16 processors per node and 4 nodes (64 procs in total), with nb=1,2,4,8 the results are all the same, and the ACE energies in the output file are ACE energy-9.83550603 ACE energy-9.85593641 ACE energy-9.86091525 ACE energy-9.86195787 ACE energy-9.86217893 ACE energy-9.86222150 but with nb=16, the result is different: ACE energy-9.57847448 ACE energy-7.73229301 == calculation='scf', outdir='./', prefix="Si8", verbosity='high', disk_io = "none" / ibrav=2, celldm(1)=20.526, ntyp = 1, nat=16, ecutwfc=90, input_dft = "pbe0", exx_fraction = 0.1449 exxdiv_treatment = 'gygi-baldereschi' ecutfock = 180 / electron_maxstep=600, mixing_beta=0.7, conv_thr=1.0D-8, / ATOMIC_SPECIES Si 28.0855 Si_ONCV_PBE-1.1.upf ATOMIC_POSITIONS alat Si 0 0 0 Si 0.125 0.125 0.125 Si 0.25 0.25 0 Si 0.375 0.375 0.125 Si 0.25 0 0.25 Si 0.375 0.125 0.375 Si 0 0.25 0.25 Si 0.125 0.375 0.375 Si 0.25 0.25 0.5 Si 0.375 0.375 0.625 Si 0.25 0.5 0.25 Si 0.375 0.625 0.375 Si 0.5 0.25 0.25 Si 0.625 0.375 0.375 Si 0.5 0.5 0.5 Si 0.625 0.625 0.625 K_POINTS automatic 4 4 4 0 0 0 Feng >Message: 6 >Date: Tue, 18 Jul 2017 16:59:52 +0200 >From: Paolo Giannozzi>Subject: Re: [Pw_forum] Band parallelization of EXX ACE in QE 6.1 >To: PWSCF Forum >Message-ID: >
[Pw_forum] Band parallelization of EXX ACE in QE 6.1
Dear everyone, I am trying the new EXX-ACE in QE6.1 on a system consists of 215 atoms, and 933 electrons of spin up and 932 down, gamma point only. When no parallelization level is specified in the command line, it runs well. With -nb 2, it runs much faster with the same result. But with -nb 4, it gives a wrong result in the first iteration of EXX and everything followed is wrong. Is there any restriction of the band parallelization here? Thanks. Best regards, Feng -- Dr. Feng Wu Postdoctoral researcher Department of Chemistry and Biochemistry University of California, Santa Cruz Phone: 831-459-2874 ___ Pw_forum mailing list Pw_forum@pwscf.org http://pwscf.org/mailman/listinfo/pw_forum
[Pw_forum] ld1.x produce different test results between iswitch=2/3
Thanks very much for the detailed explanation! Feng >Date: Wed, 30 Jul 2014 14:58:02 +0200 >From: Paolo Giannozzi >Subject: Re: [Pw_forum] ld1.x produce different test results between iswitch=2/3 >To: PWSCF Forum >Message-ID: <1406725082.16472.3.camel at pania.fastwebnet.it> >Content-Type: text/plain; charset="UTF-8" >If I remember correctly, the first test is performed for the >semilocal form of the pseudopotential, and thus yields by >construction the same eigenvalues; the second, with the >separable (Kleinman-Bylander) form, which may sometimes >badly fail >P. >On Tue, 2014-07-29 at 19:52 +0200, WF wrote: >> Dear everyone, >>I am trying to produce a norm-conserving pp of Cu with 3s and 3p >> included. Regardless of the quality of the pp, I met some problems in the >> consistency of tests. >> >> The input file is >> >> atom="Cu" >> dft="PBE" >> config="[Ar] 3d10 4s1 4p0" >> iswitch=3 >> >> rlderiv=2.90 >> eminld=-4.0 >> emaxld=2.0 >> deld=0.01 >> nld=3 >> / >> >> pseudotype=1 >> rho0 = 0.001 >> rcore = 2.00 >> lloc = -1 >> rcloc = 2.0 >> file_pseudopw="Cu.pz-rrkj.UPF" >> / >> 3 >> 3S 1 0 2.00 0.00 0.9 0.9 >> 3P 2 1 6.00 0.00 1.1 1.1 >> 3D 3 2 10.00 0.00 1.3 1.3 >> >> configts(1)="3s2 3p6 3d10 4s1 4p0" >> / > > And the test passed as > n l nl e AE (Ry)e PS (Ry)De AE-PS (Ry) > 1 0 3S 1( 2.00) -8.43051 -8.430510.0 > 2 1 3P 1( 6.00) -5.30241 -5.302410.0 > 3 2 3D 1(10.00) -0.37027 -0.370270.0 > 2 0 4S 1( 1.00) -0.33878 -0.33844 -0.00034 > 3 1 4P 1( 0.00) -0.05121 -0.051290.8 > > However, if I write another input file to test with this pp: > > atom = 'Cu' > dft = 'PBE' > config = '[Ar] 3d10 4s1 4p0' > iswitch = 2 > > rlderiv=2.90 > eminld=-4.0 > emaxld=2.0 > deld=0.01 > nld=3 > / > > > file_pseudo = 'Cu.pz-rrkj.UPF' > configts(1)="3s2 3p6 3d10 4s1 4p0" > ecutmin = 50 > ecutmax = 300 > decut = 50 > / > > Then everything went wrong like > n l nl e AE (Ry)e PS (Ry)De AE-PS (Ry) > 1 0 3S 1( 2.00) -8.43051 -8.430510.0 > 2 1 3P 1( 6.00) -5.30241 -6.238330.93592 > 3 2 3D 1(10.00) -0.37027 -0.20201 -0.16826 > 2 0 4S 1( 1.00) -0.33878 -0.363530.02475 > 3 1 4P 1( 0.00) -0.05121 -0.051210.0 > > Also with errors and warnings like > Warning: n=1, l=0 expected 0 nodes, found 1 > > > I am not sure what I was doing wrong here. What's the difference between > tests in pp-generation and individual ones? And which one is correct? > Thanks for the help. > > Best regards, > Feng > > > College of Chemistry and Molecule Engineering, > Peking University, > Beijing, > China > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://pwscf.org/mailman/listinfo/pw_forum College of Chemistry and Molecule Engineering, Peking University, Beijing, China
[Pw_forum] ld1.x produce different test results between iswitch=2/3
Dear everyone, I am trying to produce a norm-conserving pp of Cu with 3s and 3p included. Regardless of the quality of the pp, I met some problems in the consistency of tests. The input file is atom="Cu" dft="PBE" config="[Ar] 3d10 4s1 4p0" iswitch=3 rlderiv=2.90 eminld=-4.0 emaxld=2.0 deld=0.01 nld=3 / pseudotype=1 rho0 = 0.001 rcore = 2.00 lloc = -1 rcloc = 2.0 file_pseudopw="Cu.pz-rrkj.UPF" / 3 3S 1 0 2.00 0.00 0.9 0.9 3P 2 1 6.00 0.00 1.1 1.1 3D 3 2 10.00 0.00 1.3 1.3 configts(1)="3s2 3p6 3d10 4s1 4p0" / And the test passed as n l nl e AE (Ry)e PS (Ry)De AE-PS (Ry) 1 0 3S 1( 2.00) -8.43051 -8.430510.0 2 1 3P 1( 6.00) -5.30241 -5.302410.0 3 2 3D 1(10.00) -0.37027 -0.370270.0 2 0 4S 1( 1.00) -0.33878 -0.33844 -0.00034 3 1 4P 1( 0.00) -0.05121 -0.051290.8 However, if I write another input file to test with this pp: atom = 'Cu' dft = 'PBE' config = '[Ar] 3d10 4s1 4p0' iswitch = 2 rlderiv=2.90 eminld=-4.0 emaxld=2.0 deld=0.01 nld=3 / file_pseudo = 'Cu.pz-rrkj.UPF' configts(1)="3s2 3p6 3d10 4s1 4p0" ecutmin = 50 ecutmax = 300 decut = 50 / Then everything went wrong like n l nl e AE (Ry)e PS (Ry)De AE-PS (Ry) 1 0 3S 1( 2.00) -8.43051 -8.430510.0 2 1 3P 1( 6.00) -5.30241 -6.238330.93592 3 2 3D 1(10.00) -0.37027 -0.20201 -0.16826 2 0 4S 1( 1.00) -0.33878 -0.363530.02475 3 1 4P 1( 0.00) -0.05121 -0.051210.0 Also with errors and warnings like Warning: n=1, l=0 expected 0 nodes, found 1 I am not sure what I was doing wrong here. What's the difference between tests in pp-generation and individual ones? And which one is correct? Thanks for the help. Best regards, Feng College of Chemistry and Molecule Engineering, Peking University, Beijing, China
[Pw_forum] SCF Convergence problem in all-electron calculation of ld1.x
Dear everyone, I would like to verify the validity of a norm-conserving pp of carbon by use some different configurations as what documentation says. However, I cannot complete all-electron calculation of C with configuration "[He] 2s2 2p3", in which ld1.x shows "warning: convergence not achieved". The convergence cannot be achieved by adjusting beta and maximum iteration count from what I have tried. What can make it converge or should I just skip this step ? Thanks very much. Wu Feng, Chemistry and Molecular Engineering, Peking University
[Pw_forum] How to test pseudo potential for non-metallic element?
Thanks for your advices and I will try it later. Happy New Year! >You can just combine them with other known to work pseudopotentials and check a variety of compounds, e.g. H2O/H2S molecules/cluster, metal/semiconductor oxides. >Cheers, > axel. >Dear Wu, >To check the transferability it's good to try different electronic environments. The ld1 tests are the first thing to do. If they look ok, then try very different structures, including at least a bulk and the dimer, and compare with the literature. This will give you an idea about how transferable the pseudo is. Of course you can try molecules as well if you know what result to expect. >Happy new year! >Giovani Faccin Wu Feng, Chemistry and Molecular Engineering, Peking University
[Pw_forum] How to test pseudo potential for non-metallic element?
Dear everyone, I tried to create pseudo potential for O and S ( for some uncommon functional ) for later oxides calculation. Before used in the final model, it should be checked. However, for a metallic element, it is easy to create a model with its stable elementary substance and optimized to see its lattice constant as first step verification. However for an element exists as molecule, is it a good method to test the molecule case? Is it enough to use ld1.x test only as first step verification? Thanks. Wu Feng, Chemistry and Molecular Engineering, Peking University -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20111231/3f531769/attachment.htm
[Pw_forum] How to achieve the convergence of Ecutrho?
>Dear all, > >I've asked sevral people about the convergence criteria of Ecutrho test. They point out that the convergence will be achieved when the difference between two adjacent scf calculation is smaller than 0.1 meV. However, they are not very sure about the step length. From my view, the step length may pose huge effects on our calculation. Consider the cases of step lengths are 100 Ry and 0.1Ry respectively, the latter will be convergent immediately while the former will not be so fast. >For the resons above, I'm really confused now. And could any nice guy show me the strict judgement of the convergence for Ecutrho? Thank you very much! The convergence of Ecutrho comes from the balance of time and accuracy, so the judgment is based on your own system and there is no universal one. Typically, 1Ry~10Ry should be step length under most circumstance, as one can reach the convergence reasonably fast while the additional time cost from the more than enough Ecutrho will be in the same order of magnitude as 10%, an acceptable value. Wu Feng, Chemistry and Molecular Engineering, Peking University
[Pw_forum] A minor mistype in INPUT_PW doc?
Dear QE developers, Today I find a word seems mistyped in Doc/INPUT_PW.html and INPUT_PW.txt in the k-points description part. Syntax: K_POINTS tpiba | crystal | tbiba_b | crystal_b Is the "tbiba_b" should be tpiba_b as used elsewhere? Thanks. -- Wu Feng Chemistry and Molecular Engineering, Peking University -- -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20111207/fd79a558/attachment.htm
[Pw_forum] band structure for a 72 C atom supercell
> Dear Wu, > Thanks for your reply. Is there any other executable that does this?what about Gnuplot? does it follow the same procedure or is different? > Thanks again > Elie The band information can be found in the stdout of pw.x in text. Any graph tool include gnuplot and xmgrace should be OK to display them, however it may be necessary to rearrange the data to an usable format by yourself. -- Wu Feng Chemistry and Molecular Engineering, Peking University --
[Pw_forum] band structure for a 72 C atom supercell
At Sat, 3 Dec 2011 13:53:03 +, Elie M writes: >Dear all, >I computed the electronic band structure of a 72 Carbon atom supercell but the result shows a lot of wiggles in the ? energies? Do I have to use more points or might it be due to something else? >You will find the ps plot attached >Regards >Elie MoujaesUnIVERISTY od NottinghamNG7 2RDUK In my experience the band structure plotted by plotband.x may give unphysical band crossing. You can just plot the raw data in the original order to avoid this, however band-crossing will be totally lost. -- Wu Feng Chemistry and Molecular Engineering, Peking University --
[Pw_forum] Does total force include dispersion force?
Hello everyone, I am doing a vc-relax calculation with PBE+D ( with .london. = true and PBE functional). In the output, an additional part called "Dispersion forces acting on atoms" is displayed after "Total forces acting on atoms" after each BFGS cycle. Do the values in total force part already include dispersion forces ? I am not sure with what I read from the codes. Thanks very much. Wu F, College of Chemistry and Molecular Engineering, Peking University. -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/2024/4c6a96f5/attachment.htm
[Pw_forum] What kind of calculation needs wavefunction files created by pw.x ?
> Actually, wavefunctions are not needed for DOS calculations, but they are for projected-DOS ones. For calculations with bands.x, you will need wavefunctions only in conjunction with the lsym=.TRUE. option. Other standard calculations that require wavefunctions are, for instance, LDOS (and STM images), Wannier function analysis and maybe others. Also, sometimes you may want to use the final charge and wavefunctions as a starting guess for a successive calculation with pw.x. Thank you very much for these detailed explanation. > ? Gabriele Sclauzero, EPFL SB ITP CSEA >PH H2 462, Station 3, CH-1015 Lausanne > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum Wu F, College of Chemistry and Molecular Engineering, Peking University.
[Pw_forum] What kind of calculation needs wavefunction files created by pw.x ?
Hello everyone, From what I haved learned, pw.x generates wavefunction files under temporary directory, and can be collected to output directory by options wf_collect = .true. . It can be controlled by disk_io option to decrease the I/O. However I don't really know whether should I keep this files to do later calculation like DOS and bands or not. What kind of calculation needs wavefunction files created by pw.x ? Thanks. Wu F, College of Chemistry and Molecular Engineering, Peking University. -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/2014/a0649e0c/attachment.htm
[Pw_forum] How to show bands above fermi level in PDOS?
Hello everyone, I am calculating PDOS of an insulator. In the initial SCF, keyword 'nbnd' is not specified. A NSCF calculation, still without 'nbnd' keyword, is done to get Fermi level. And in the later PDOS calculation, no band above Fermi level is showed. If I need to show some bands above Fermi level, from where should I restart my calculation? Is it necessary to rerun SCF with a large 'nbnd' or there are some quicker way to do it ? Thanks. Wu F, College of Chemistry and Molecular Engineering, Peking University. -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/2005/148f1fa6/attachment.htm
[Pw_forum] c_bands: X eigenvalues not converged
Hello everyone, I am running a bands calculation based on a scf calculation already done without problem. But after several days the output keeps prompting "c_bands: X eigenvalues not converged", where X = 3~8, and there are already approximately 300 lines of such messages. Is it just in the process or it meets problem and I must kill it? Thanks very much. -- Wu, F College of Chemistry and Molecular Engineering Peking University --- -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20111030/0e2a210a/attachment.htm
[Pw_forum] Problem of the energy of orbital in ld1.x with
On Sep 26, 2011, at 8:30, Paolo Giannozzi wrote: > energy = 0.00 => "use energy of the bound state" > energy /=0.00 => "use specified energy for pseudization" > It is clumsy and unclear but it was done that way and now it cannot > be changed > without breaking existing data. Anyway, it is documented. > In the latter case, the pseudo-wavefunction is not used for projection: > it is not guaranteed to be a bund state Yes it is documented in atomic package related guide, but I have not understand it clearly until you explain this. Thanks very much for your patience guidance. Best regards, Wu --- F, Wu College of Chemistry and Molecular Engineering Peking University --
[Pw_forum] Problem of the energy of orbital in ld1.x with
Thanks for your advices. How can I get bound 5P states ? In my opinion a state has energy lower than 0 is bound, but regardless what the 5P energy in ld1.x input file is ( I test the range from -0.1 to 1.50 ), no 5P orbital is found in PDOS calculation, which only appear when 5P energy is set to 0.00 . On Sep 25, 2011, at 18:31:16, Paolo Giannozzi wrote: >> my problem here is [...] how to show 5P in later PDOS calculation >> >you need atomic wavefunctions for a bound 5P state in the pseudopotential file > >P. >--- >Paolo Giannozzi, Dept of Chemistry, Univ. Udine, via delle Scienze 208, 33100 Udine, Italy Phone +39-0432-558216, fax +39-0432-558222 --- F, Wu College of Chemistry and Molecular Engineering Peking University --
[Pw_forum] Problem of the energy of orbital in ld1.x with PDOS
The FHI file seems doesn't contain all information necessary and hard to convert. However, my problem here is not about the quality of the PP, but how to show 5P in later PDOS calculation. Currently PDOS of Ru metal with PP that energy of 5P is not 0.00 will give result that the PDOS of 5P does not exists. On 24 Sep, 2011 ,at 19:56:20, GAO Zhe wrote: >Even though I cannot give you any suggestion to generate a better NCPP for Ru, you can try to convert a fhi NCPP to UPF by upftools/fhi2upf.x. Here: http://www.sas.upenn.edu/rappegroup/htdocs/Research/psp_gga.html#Ru3, you can find both of the fhi file and parameters. > > >-- >GAO Zhe >CMC Lab, MSE, SNU, Seoul, S.Korea > > > > >At 2011-09-24 17:45:49,WF wrote: >>Hello everyone, >> I recently do a vc-relax calculation of Ru metal of PBE >>norm-conserving PP. >> >>The PP input is like this, which is slightly modified from that in >>atomic_doc: >>-- >> >> title='Ru', >> zed=44.0, >> rel=1, >> config='[Kr] 4d6.0 5s2.0 5p0', >> iswitch=3, >> dft='pbe' >> / >> >> pseudotype=2, >> file_pseudopw='Ru.pbe-rrkjnc.UPF', >> author='TM', >> lloc=0, >> nlcc=.true., >> rcore = 0.9, >> / >>3 >>5P 2 1 0.00 -0.099963 2.40 2.40 0.0 4D 3 2 6.00 0.00 1.90 >>2.40 0.0 5S 1 0 2.00 0.00 2.30 2.40 0.0 >>--- >>--- >> >>If the 5P energy is set to 0.00 ( which means energy of all-electron >>calculation ), the a axis of result is identical to that when the >>energy is set to -0.09963 ( the all electron result copied from ld1.x output ). >>However, in .UPF file, 2 wavefunctions and 3 projectors are generated >>if the energy is 0.00, all of 4D, 5S and 5P will appear in PDOS ( by >>projwfc.x ), but only 2 projectors are obtained otherwise and 5P will >>not appear in further PDOS calculation. I don't know how to make PDOS >>of all of 4D, 5S and 5P if the energy is not 0.00. Can anyone give some advices ? Thanks. >> >>Best regards, >> >>Wu F >> >>--- >>F, Wu >>College of Chemistry and Molecular Engineering Peking University >>-- >> > --- F, Wu College of Chemistry and Molecular Engineering Peking University -- -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110925/647781e1/attachment.htm
[Pw_forum] Problem of the energy of orbital in ld1.x with PDOS
Hello everyone, I recently do a vc-relax calculation of Ru metal of PBE norm-conserving PP. The PP input is like this, which is slightly modified from that in atomic_doc: -- title='Ru', zed=44.0, rel=1, config='[Kr] 4d6.0 5s2.0 5p0', iswitch=3, dft='pbe' / pseudotype=2, file_pseudopw='Ru.pbe-rrkjnc.UPF', author='TM', lloc=0, nlcc=.true., rcore = 0.9, / 3 5P 2 1 0.00 -0.099963 2.40 2.40 0.0 4D 3 2 6.00 0.00 1.90 2.40 0.0 5S 1 0 2.00 0.00 2.30 2.40 0.0 -- If the 5P energy is set to 0.00 ( which means energy of all-electron calculation ), the a axis of result is identical to that when the energy is set to -0.09963 ( the all electron result copied from ld1.x output ). However, in .UPF file, 2 wavefunctions and 3 projectors are generated if the energy is 0.00, all of 4D, 5S and 5P will appear in PDOS ( by projwfc.x ), but only 2 projectors are obtained otherwise and 5P will not appear in further PDOS calculation. I don't know how to make PDOS of all of 4D, 5S and 5P if the energy is not 0.00. Can anyone give some advices ? Thanks. Best regards, Wu F --- F, Wu College of Chemistry and Molecular Engineering Peking University --
[Pw_forum] Error: fft order too large
Hello everyone, I am calculating a surface slab model of MoS2. My parameter works well with 5~11 layers. But with 13 layers ( 39 atoms total ), it gives %% from good_fft_order : error # 2241 fft order too large %% Decreasing Ecutwfc/Ecutrho can eliminate these problem. But these Ecutwfc/Ecutrho is determined by convergence test of bulk lattice parameter test, so I would like to find a way to solve this problem without decreasing Ecutwfc/Ecutrho. Is there any suggestion ? My input file is like this: ( all 5~11layers input files are just same except atoms related part ) == calculation = 'relax' restart_mode = 'from_scratch' tstress = .true. tprnfor = .true. outdir = './out' wfcdir = './wfc' prefix = 'MoS2' / ntyp = 2 nat = 39 ibrav = 4 celldm(1) = 6.108157 celldm(3) = 40.745230 ecutwfc = 100.0 ecutrho = 800.0 london = .false. / diagonalization = 'cg' mixing_mode = 'plain' mixing_beta = 0.7 conv_thr = 1.00d-08 / ion_dynamics = 'bfgs' ion_positions = 'default' / ATOMIC_SPECIES S 32.07 S.vdwDF-tmus-tmc.UPF Mo 95.94 Mo.vdwDF-rrkjus-tmc.UPF (Atomic position is negelected for it is too long) K_POINTS automatic 8 8 1 1 1 1 = Best Regards, F, Wu --- F, Wu College of Chemistry and Molecular Engineering Peking University -- -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110915/862b0759/attachment.htm
[Pw_forum] problems with vc-relax
Is it good enough to relax an arbitrary system with only gamma k-points? I thought it is necessary to do some test with k-points as this effect change from system to system. In my experience, more k-points is necessary for a metal-like system, i.e. graphene. Best regards, Wu, F > Date: Thu, 8 Sep 2011 11:12:24 -0400 > From: xijunw at gmail.com > To: pw_forum at pwscf.org > Subject: Re: [Pw_forum] problems with vc-relax > > On Thu, Sep 8, 2011 at 10:26 AM, Elie Moujaes > wrote: > > Hello Xijun, > > Hello Elie, > > > I did not send the e-mail twice intentionally. There was a problem with my > > e-mail and it kept giving me an error about problems with Windows hotmai > > hence why it was sent twice without me knowing except now when I opened my > > e-mail so I apologize for that. Also I wont include blank lines in the input > > file either. Back to my calculations, I have restarted the calculations, it > > went fine at first but then I got the error: > > from scale_h: not enough memory allocated for radial FFT; try restarting > > with a larger cell_factor. > > Check the manual for definition of cell_factor. The default value is > 1.2, you could try a larger value. > > What's your new K_POINTS setting? Since you are relax the system, I > suggest you start from: > > K_POINTS {gamma} > > This will far faster. and need far less memory. > > > Another simple question: I realized when I restarted everything that the > > positions of the atoms were the initial ones even before starting the > > relaxation process where as the k-points used were the NEW ones. Is that > > normal? or should I have substituted the the C atoms positions by the most > > recent ones then restart again? > > Yes, restart again with NEW ATOMIC_POSITIONS and "restart_mode= 'from_scratch' " > > Regards, > - Xijun > > > Regards > > Elie > > > >> Date: Wed, 7 Sep 2011 16:20:49 -0400 > >> From: xijunw at gmail.com > >> To: pw_forum at pwscf.org > >> Subject: Re: [Pw_forum] problems with vc-relax > >> > >> Hi, Elie, > >> > >> Please do not repeat sending the same message to the mail list. Also, > >> people will be more happy to help you if you keep your mail neat by > >> removing all the unnecessary blank lines in the input file. > >> > >> Regards, > >> Xijun > >> > >> On Wed, Sep 7, 2011 at 3:28 PM, Elie Moujaes > >> wrote: > >> > Dear all, > >> > > >> > > >> > > >> > I am relaxing a 60 atom supercell (graphene grain boundary) to make the > >> > total force on atoms mainly 0.001 Ry/au. However the vc-relax stopped > >> > after > >> > 4 days of execution and without reaching the desired force threshold. > >> > The > >> > following output was obtained : > >> > > >> > > >> > > >> > 1st relaxation process: > >> > > >> > > >> > > >> > Total force 0.091 Pressure =- 114.1 kbar > >> > > >> > > >> > > >> > 2nd relaxation process: > >> > > >> > > >> > > >> > total force 0.1 Pressure = -55 kbar > >> > > >> > > >> > > >> > (new enthalpy < old enthalpy) > >> > > >> > > >> > > >> > 3rd process: > >> > > >> > > >> > > >> > total force = 0.058 pressure = -11 kbar > >> > > >> > > >> > > >> > Then in the 4th process, calculations stopped. Here are the last few > >> > lines > >> > of the output: > >> > > >> > > >> > > >> > iteration # 2 ecut=36.75 Ry beta=0.30 > >> > Davidson diagonalization with overlap > >> > c_bands: 5 eigenvalues not converged > >> > c_bands: 1 eigenvalues not converged > >> > c_bands: 3 eigenvalues not converged > >> > c_bands: 1 eigenvalues not converged > >> > c_bands: 4 eigenvalues not converged > >> > c_bands: 4 eigenvalues not converged > >> > c_bands: 4 eigenvalues not converged > >> > c_bands: 4 eigenvalues not converged > >> > c_bands: 5 eigenvalues not converged > >> > c_bands: 5 eigenvalues not converged > >> > c_bands: 2 eigenvalues not converged > >> > c_bands: 2 eigenvalues not converged > >> > ethr = 2.90E-04, avg # of iterations = 17.6 > >> > > >> > negative rho (up, down): 0.132E-01 0.000E+00 > >> > > >> > total cpu time spent up to now is 382283.13 secs > >> > > >> > total energy =-676.77947904 Ry > >> > Harris-Foulkes estimate =-676.82846513 Ry > >> > estimated scf accuracy< 0.17363554 Ry > >> > > >> > iteration # 3 ecut=36.75 Ry beta=0.30 > >> > Davidson diagonalization with overlap > >> > c_bands: 1 eigenvalues not converged > >> > c_bands: 1 eigenvalues not converged > >> > c_bands: 2 eigenvalues not converged > >> > c_bands: 3 eigenvalues not converged > >> > > >> > > >> > > >> > The input of the vc-relax is: > >> > > >> > > >> > > >> > > >> > > >> > prefix='GBphonon', > >> > > >> > calculation='vc-relax', > >> > > >> > restart_mode='from_scratch', > >> > > >> > tstress=.true., > >> > > >> > tprnfor=.true, > >> > > >> > pseudo_dir = > >> >
[Pw_forum] Is it necessary to clear outdir / wfcdir before calculation?
Thanks for your advices, I will remove the output if it is not necessary. I am sorry that I don't notice that the title contains a newline mark of which only the first line is copied in my message sent, so it doesn't work properly, i.e. no previous message or affiliation. Best regards, Wu F --- F, Wu College of Chemistry and Molecular Engineering Peking University --
[Pw_forum] Is it necessary to clear outdir / wfcdir
In my calculation restart_mode="from_scratch" is always used. But I find if a calculation ends with error, it will give different errors depends on the whether the outdir is empty or not. So I wonder whether it affect the calculation results.
[Pw_forum] Is it necessary to clear outdir / wfcdir before calculation?
Hello everyone, Recently I make a small bash script to run a lot of small pw.x calculation one by one. But all of them use same outdir and wfcdir. They are temporary directories and I don't need them ( I only need scf energy, so stdout is enough ). Is it necessary to delete them before starting a new calculation ? Because I have done a lot of calculation, I would like to verify whether this will affect the results to determine whether I should modify the script to restart my calculation again, or just use these result. Best regards, Wu F --- F, Wu College of Chemistry and Molecular Engineering Peking University -- -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110908/e9fad8f3/attachment.htm
[Pw_forum] Problem with Bi2Se3 SCF converge
Dear Mohnish: Thanks for our advices. I tried beta = 0.1 and 0.2, and those message is postponed to 100+ iteration, but the convergence still cannot be achieved. On Fri, Sep 2, 2011 at 11:45 AM, mohnish pandey wrote: > Try reducing "beta" >On Wed, Aug 31, 2011 at 8:07 AM, WF wrote: > Hello everyone, > > I am doing calculation of Bi2Se3 with vdW-DF functional, but > it really hard to converge. After several iterations it keeps to > prompt things > like: > > ** ** > > iteration # 58 ecut=60.00 Ry beta=0.70 > > CG style diagonalization > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 2 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 2 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 1 eigenvalues not converged > > c_bands: 2 eigenvalues not converged > > c_bands: 2 eigenvalues not converged > > c_bands: 2 eigenvalues not converged > > c_bands: 2 eigenvalues not converged > > > > and will not converge in 1000 iteration. > > ** ** > > But the calculation of LDA and PBE can converge, however. All > pseudo-potential is generated by ld1.x with same parameters except > functional. > > ** ** > > I have tried to change some parameter in SCF but it doesn?t > seem to be helpful. Is there any suggestion for this problem? > > ** ** > > Thanks. > > ** ** > > F,Wu > > ** ** > > ___ > Pw_forum mailing list > Pw_forum at pwscf.org > http://www.democritos.it/mailman/listinfo/pw_forum > > - Regards --- F, Wu College of Chemistry and Molecular Engineering Peking University --
[Pw_forum] vdW functional in Quantum-Espresso
There are large changes in vdW functional part between version 4.3 and 4.3.1 or later. First, the vdW functional is put in GGA correlation part in funct.f90 in 4.3, but in an independent part (non-local part) in funct.f90 in 4.3.1. Second, the ld1.x refuse to generate pseudo-potential for functional containing vdW part like vdW-DF in 4.3.1 for it is non-local but it is possible to generate something in 4.3. So I have some question about vdW functional. 1. In 4.3, does vdW-DF, or "sla+pw+rpb+vdw" as 1/4/4/9 imply the original GGA correlation in revPBE, "pbc" here, is used? As 4.3.1 allow user to specify GGA correlation and vdW correction simultaneously, and vdW-DF in that version is 1/4/4/0/1, is it reasonable to add GGA correlation in the functional and perform further calculation? I note in literature it is said GGA correlation is small so it is neglected. 2. What is the difference of vdW functional correction in vdW-DF1 and vdW-DF2? 3. If it is impossible to get a pseudo-potential of vdW-DF , what functional in pseudo-potential should I specify to perform calculation later? What is the pp generated by 4.3 actually and can they be used in further calculation? Thanks. -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110824/e6c3ec4b/attachment.htm