[Pw_forum] PHonon versions 5.0.2/5.0.3 can give very different frequencies as compared to older versions and the latest SVN builds
In the event that someone stumbles across this thread in the future and looks for an answer by subject line, I just wanted to note here that this issue has been resolved by the new 5.0.3 patch that Paolo released this morning. Best, Brad Malone Harvard University -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130506/95a8f654/attachment.html
[Pw_forum] PHonon versions 5.0.2/5.0.3 can give very different frequencies as compared to older versions and the latest SVN builds
Hi everyone, I recently came across some phonon softening in a system calculated with the 5.0.2 version of QE. This came as a surprise because, in addition to having some physical reasons for not believing it, it was in contrast to previous calculations I had done some time back on the same system (and input files) with QE 4.3.2. The deviations with respect to QE version over a 3x3x3 grid of q vectors in this system amounted to a maximal error of 13.5 cm^-1 for a particular q vector. This previously mentioned system was fairly large, so I tried to find simpler systems to run so that I could narrow down where the problem was coming from. I found that many simple systems (like cubic diamond) exhibited essentially zero error regardless of version. However, in guessing that the problem could possibly be related to systems with hexagonal symmetry, I chose to test the hexagonal diamond phase of Ge (4 atoms/cell) and found that over a 3x3x3 grid of q vectors, I got deviations on the order of 95 cm^-1 for some q vectors. I actually do not think the problem is limited to hexagonal symmetry systems however, as I've seen some deviations in other orthorhombic systems (although they have been more minor). Using the hexagonal diamond system, I tested a wide variety of QE releases and SVN versions. The problem occurs for the more recent releases of 5.0.2 and 5.0.3 (i.e., the patch doesn't fix the problem). The latest SVN version is consistent with 5.0.1 and 5.0.0 as well as older versions tested (4.3.2 and 4.2.1). I'm not sure the exact SVN version where the problem first occurred, but rev 9724 is the first commit which gives the "correct" phonon frequencies. Rev 9724 differs from 9723 as follows: login1$ svn -r 9723:9724 diff > Index: PHonon/PH/sgam_ph.f90 > === > --- PHonon/PH/sgam_ph.f90 (revision 9723) > +++ PHonon/PH/sgam_ph.f90 (revision 9724) > @@ -161,7 +161,8 @@ > sym(irot) = sym(irot) .and. (abs(raq(ipol)-aq(ipol)) < 1.0d-5) > enddo > endif > - if (.not.minus_q) then > +! if (.not.minus_q) then > + if (sym(irot).and..not.minus_q) then > raq = - raq > minus_q = eqvect (raq, aq, zero, accep) > endif Anyway, hope these tests are useful. Unless I've made a mistake somewhere, I think it suggests that it's extremely dangerous to do phonon calculations with 5.0.2 or 5.0.3 and you should either stick to 5.0.1 or svn revision 9724 or later. Best, Brad Malone Harvard University -- next part -- An HTML attachment was scrubbed... URL: http://pwscf.org/pipermail/pw_forum/attachments/20130503/c9347638/attachment.html
[Pw_forum] Pw_forum Digest, Vol 48, Issue 68
Dear Stefano, I appreciate your additional remarks and Claudia's thesis. They are both very helpful. I'll continue to look into this possible Kohn anomaly and its relationship with the lattice instability as the pressure is lowered. Thanks again, Brad UC Berkeley > > Dear Brad & Nicola, > > I was going to reply to Brad, when I noticed that Nicola already did so, > and quite appropriately. Let me just elaborate a little bit on Nicola's > remarks. > > The reason why Kohn anomalies require so many k-points to be properly > captured on a computer is the same why they are so sensitive to temperature > in nature. They are due to different portions of the Fermi surface to be > quasi-parallel (i.e. connected by a same q vector, the "nesting" vector). By > the way, this is why Kohn anomalies are more important on low dimensions: > the lower the dimension, the easier it is to have nesting. When this occurs, > perturbations with the periodicity of the nesting vector will be strongly > screened, hence phonons with that periodicity will be soft or "quasi soft". > When the temperature increases, the Fermi surfaces becomes "blurred", and > the very concept of nesting breaks down. Computationally, in a metal the > energy smearing (Gaussian or other) plays the role of an effective > temperature. The smaller the smearing, the larger the number of k-points > necessary to sample the Brillouin zone to an energy resolution compatible > with the smearing. > > An early paper studying a system where these effects show dramatically is: > > Claudia Bungaro et al. PRL 77, 2491 (1996) > http://link.aps.org/doi/10.1103/PhysRevLett.77.2491 > You may also find Claudia's PhD thesis worth some attention: > http://www.sissa.it/cm/thesis/1995/bungaro.ps.gz > > Hope this may help. > > Stefano > > > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110627/42a47d76/attachment.htm
[Pw_forum] Convergence of imaginary modes
Dear Nicola, Thanks so much for your suggestion; I was not aware of this characteristic of Kohn anomalies. I will look into some of those papers (a quick glance at them already suggests that they may be useful). My smearing is pretty "standard" for metallic systems at 0.03 Ry, and so I will look into the possible presence of a Kohn anomaly in the system. Best, Brad UC Berkeley > Dear Brad, > > > if you have a Kohn anomaly it's not unusual to require that high > of a k-point sampling - comparable numbers are needed to converge > some of the optical modes in graphene/graphite (do check some of > the early papers of Mauri/Lazzeri). > > I haven't heard of other cases where such high-sampling is needed > for phonons, unless your system was a metal, and you were not using > a smearing scheme (or way too small a smearing). > >nicola > > > -- > -- > Prof Nicola MarzariDepartment of MaterialsUniversity of Oxford > Chair of Materials Modelling Director, Materials Modelling Laboratory > nicola.marzari at materials.ox.ac.uk http://mml.materials.ox.ac.uk/NM > > > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110626/d270c2a0/attachment.htm
[Pw_forum] Convergence of imaginary modes
Hello QE users, I am calculating the lattice dynamical properties of a high-pressure phase. As has been documented in this forum, phonon frequencies can be quite sensitive to the various other parameter of the calculation (wavefunction cutoff, BZ sampling, etc.) and it's important to test this explicitly and not rely on the parameters one finds converged for say, a total-energy SCF calculation. In the testing of my phonon frequencies with respect to the density of the k-point sampling in the SCF calculation prior to the phonon calculation, I decided to test a uniform grid of q-points rather than simply the Gamma phonon since I don't know of any a priori reason to expect that all phonon frequencies converge identically (although in my prior experience this is approximately true). I found that almost all frequencies are converged to within about 1% or so at a "small" kgrid sampling of about 40x40x40. However, one mode is wildly unconverged at this point, and differences on the order of 100s of cm^1 can be found by going to a kgrid sampling of 80x80x80. So I just kept increasing the kpoint sampling until I converged this mode. The mode didn't fully converge until 200x200x200, although it is pretty close at 160x160x160.The mode finally converged to a negative value (although was positive until the kpoint sampling was in the triple digits in each direction). This negative mode signals an instability of the structure along the path corresponding to the phonon displacement pattern. That's all well and good. I know that this structure is thermodynamically unstable at pressures in the ballpark of the calculation, and so if it's dynamically unstable too no problem. Since the structure becomes more thermodynamically stable at higher presures, I increased the pressure and tested the phonon frequencies again. I found that the negative mode goes positive, and increases in frequency as the pressure is increased (as one might guess). However, this mode still depends very sensitively on the kpoint sampling and doesn't get close to full convergence until about 160x160x160 similarly to the calculation at lower pressures. In previous lattice dynamical calculations, I've always been blessed with positive frequencies (at least those that can't be ASRed away) and I'm curious as to whether the more sensitive convergence of imaginary modes (and those that are close to being unstable) is a property that others have found in their calculations or whether it is something unusual with the system that I am studying. Thanks, Brad Malone UC Berkeley -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20110626/514877af/attachment.htm
[Pw_forum] Details on "third order derivatives not implemented with GGA" error
Lorenzo and Paolo, Thanks for the replies, I appreciate it. I'll contact Tobias to inquire about the status. Best, Brad UC Berkeley On Tue, Nov 9, 2010 at 9:01 PM, Brad Malone wrote: > >> >> FYI. >> There is NO truncation on the mailing list, we got all what you sent. >> You gmail did it. Click Show Quote Text for full email. >> >> > That doesn't seem to be completely true. Check out the posts here > http://www.democritos.it/pipermail/pw_forum/2010-November/date.html . What > is seen on the digest is different than what is stored online at the site > above for some reason. > > > > Best, > Brad > UC Berkeley > > > > > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20101110/014a4112/attachment.htm
[Pw_forum] Details on "third order derivatives not implemented with GGA" error
No idea why it is being truncated, so this time I'll start the email after the error message ---
[Pw_forum] Details on "third order derivatives not implemented with GGA" error
It appears that my last email was severely truncated on the mailing list (even though it looks fine in my outbox). Below is what is should have been: --- Hi, I am seeking more information on the error > from phq_setup : error # 1 > third order derivatives not implemented with GGA
[Pw_forum] Details on "third order derivatives not implemented with GGA" error
Hi, I am seeking more information on the error > from phq_setup : error # 1 > third order derivatives not implemented with GGA
[Pw_forum] about the program style of pwscf
The book 'FORTRAN 90/95 for Scientists and Engineers' goes through basic FORTRAN 90 syntax, and is pretty good. There's a new book out for FORTRAN 95/2003, but the differences should be minor. Best, Brad Malone UC Berkeley On Wed, Aug 11, 2010 at 3:19 AM, wrote: > > first thank you for you reply in details, yes ,just like you said, the > pwscf > is written mostly in standard Fortran90, and what I want to is just to > learn > this standard style of program. > 2010/8/11 Gabriele Sclauzero > > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100811/26645d79/attachment.htm
[Pw_forum] What happens at REALLY large ectuwfc?
>start with 'random' or 'atomic+random' initial wavefunctions >(startingwfc='...'). I didn't find anything wrong with very >high cutoffs. Occasionally you can end up in the wrong ground >state, though, especially in highly symmetric cases like the >Si example you sent. Hi Paolo. I tried startingwfc='random' and the eigenvalues are now essentially the same as the 200 Ry calculation. k = 0. 0. 0. (** PWs) bands (ev): -5.1746 7.0369 7.0369 7.0369 !total energy = -14.59208472 Ry Harris-Foulkes estimate = -14.59208472 Ry estimated scf accuracy<3.8E-11 Ry The total energy is the sum of the following terms: one-electron contribution = 5.58227211 Ry hartree contribution = 1.67255887 Ry xc contribution =-5.04795092 Ry ewald contribution= -16.79896478 Ry Fock energy 1 = 0. Ry Fock energy 2 = 0. Ry Half Fock energy 2= 0. Ry Thanks, Brad On Sun, Jan 24, 2010 at 2:38 PM, Brad Malone wrote: > >are you using the latest cvs version? apparently there is a problem > >with the new symmetrization algorithm that will be fixed ASAP. > > The results I posted are from espresso-4.0.5, although I originally saw > this problem with espresso-4.1.1 in a different system (AlAs on a 2x2x2 > shifted grid). > > As for what Lorenzo said, it makes sense with what I'm seeing. The energy > breakdowns for the 200 Ry and the 2000 Ry cases are shown below: > > For 200 Ry: > > !total energy = -14.59208467 Ry > Harris-Foulkes estimate = -14.59208467 Ry > estimated scf accuracy<3.9E-11 Ry > > The total energy is the sum of the following terms: > > one-electron contribution = 5.58227319 Ry > hartree contribution = 1.67255719 Ry > xc contribution =-5.04795028 Ry > ewald contribution= -16.79896478 Ry > Fock energy 1 = 0. Ry > Fock energy 2 = 0. Ry > Half Fock energy 2= 0. Ry > - > For 2000 Ry: > - > !total energy = -14.11008918 Ry > Harris-Foulkes estimate = -14.11008918 Ry > estimated scf accuracy<1.0E-11 Ry > > The total energy is the sum of the following terms: > > one-electron contribution = 6.07256114 Ry > hartree contribution = 1.58024935 Ry > xc contribution =-4.96393489 Ry > ewald contribution= -16.79896478 Ry > Fock energy 1 = 0. Ry > Fock energy 2 = 0. Ry > Half Fock energy 2= 0. Ry > --- > > So you can see the one-electron contribution going up and the hartree > contribution going down as Lorenzo as argued. > > Brad > -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100125/4075c9ae/attachment.htm
[Pw_forum] What happens at REALLY large ectuwfc?
>are you using the latest cvs version? apparently there is a problem >with the new symmetrization algorithm that will be fixed ASAP. The results I posted are from espresso-4.0.5, although I originally saw this problem with espresso-4.1.1 in a different system (AlAs on a 2x2x2 shifted grid). As for what Lorenzo said, it makes sense with what I'm seeing. The energy breakdowns for the 200 Ry and the 2000 Ry cases are shown below: For 200 Ry: !total energy = -14.59208467 Ry Harris-Foulkes estimate = -14.59208467 Ry estimated scf accuracy<3.9E-11 Ry The total energy is the sum of the following terms: one-electron contribution = 5.58227319 Ry hartree contribution = 1.67255719 Ry xc contribution =-5.04795028 Ry ewald contribution= -16.79896478 Ry Fock energy 1 = 0. Ry Fock energy 2 = 0. Ry Half Fock energy 2= 0. Ry - For 2000 Ry: - !total energy = -14.11008918 Ry Harris-Foulkes estimate = -14.11008918 Ry estimated scf accuracy<1.0E-11 Ry The total energy is the sum of the following terms: one-electron contribution = 6.07256114 Ry hartree contribution = 1.58024935 Ry xc contribution =-4.96393489 Ry ewald contribution= -16.79896478 Ry Fock energy 1 = 0. Ry Fock energy 2 = 0. Ry Half Fock energy 2= 0. Ry --- So you can see the one-electron contribution going up and the hartree contribution going down as Lorenzo as argued. Brad -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100124/059131a8/attachment.htm
[Pw_forum] What happens at REALLY large ectuwfc?
I happened to stumble upon something today that I thought was unusual. The problem is illustrated in a simple example: if you take the following basic cubic Si input file and, say, the Si.pz-vbc.UPF pseudopotential from the QE website and do a test for convergence with respect to ectutwfc. If we look at the 4 bands that are calculated we see the following results: At ecutwfc=40 Ry: -5.1740 7.0370 7.0370 7.0370 ecutwfc=100Ry: -5.1746 7.0369 7.0369 7.0369 ecutwfc=200Ry:-5.1746 7.0369 7.0369 7.0369 Now try ecutwfc=2000Ry: -5.2862 6.9066 6.9066 10.2399 Now why do the values change? If I look at the output file for the 2000 Ry case I see that there is a negative starting charge: Initial potential from superposition of free atoms Check: negative starting charge= -0.057614 But still, why is this? The usual answer for a negative starting charge is to increase the wavefunction cutoff, although I suspect that's not the problem in this case. So besides the using of a 2000 Ry cutoff for silicon, what else is wrong here? Best, Brad UC Berkeley prefix = 'si' calculation = 'scf' restart_mode = 'from_scratch' wf_collect = .false. tstress = .true. tprnfor = .true. outdir = './' wfcdir = './' pseudo_dir = './' / ibrav = 0 celldm(1) = 10.2612 nat = 2 ntyp = 1 nbnd = 4 ecutwfc = 2000.0 / electron_maxstep = 100 conv_thr = 1.0d-10 mixing_beta = 0.7 diago_full_acc = .true. / CELL_PARAMETERS cubic 0.0 0.5 0.5 0.5 0.0 0.5 0.5 0.5 0.0 ATOMIC_SPECIES Si 28.086 Si.pz-vbc.UPF ATOMIC_POSITIONS crystal Si -0.12500 -0.12500 -0.12500 Si 0.12500 0.12500 0.12500 K_POINTS automatic 1 1 1 0 0 0 -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20100122/728285a7/attachment.htm
[Pw_forum] [Fwd: diagonalization failure (david, cg) for large numbers of bands]
Joseph, I've never tried computing quite that many bands (only gone up to about 1200 bands) or with that many atoms per cell. However, I've never not been able to get the bands to converge using both 'cg' diagonalization and also with increasing the cutoff. You mentioned that you have tried increasing the cutoff, but perhaps you haven't gone high enough? Here is a quote from Paolo that might be relevant in your situation: the algorithm used in PWscf is based on the assumption that > the number of Kohn-Sham states (aka bands) is much smaller > than the number of basis functions (i.e. plane waves). If this > assumption doesn't hold (and it doesn't if you calculate 500 > bands in fcc Si unless you use a very large cutoff) you may > run into trouble. Moreover iterative diagonalization becomes > slower and more memory-consuming than conventional > diagonalization. If you really need to do that, you need to > modify the code > > Best, Brad Malone UC Berkeley -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20091201/4a3b1540/attachment.htm
[Pw_forum] Calculation of Raman tensor/intensity for material with indirect band overlap
As a follow up on my previous post, I let the calculation run to completion on a less dense k-grid and saw the following output: ik 2 ibnd 17 linter: root not converged 0.748E+35 > kpoint 2 ibnd 17 pcgreen: root not converged 0.421E+23 > Non-scf u_k: avg # of iterations =137.0 > Non-scf Du_k: avg # of iterations =200.0 > > Dielectric constant from finite-differences > > ( NaN NaN NaN ) > ( NaN NaN NaN ) > ( NaN NaN NaN ) > > Computing Second order response > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > > > iter # 1 av.it.: 147.2 > thresh= 0.100E-01 alpha_mix = 0.100 |ddv_scf|^2 = NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > kpoint 2 ibnd 17 pcgreen: root not converged NaN > > > iter # 2 av.it.: 200.0 > thresh= 0.100E-01 alpha_mix = 0.100 |ddv_scf|^2 = NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > kpoint 1 ibnd 17 pcgreen: root not converged NaN > > > and so on. The code crashes at the end with the error: %% > from broyden : error # 1 > factorization > % > > but obviously the problem occurs well before this crash. Again, I am trying to calculate the Raman intensity for a material with a small indirect overlap. Should this not be possible? I know that DFPT is applicable (without modification) only for systems with a clear distinction between occupied and valence states (in order to clearly specify what the projectors over the occupied and unoccupied manifolds I guess?). For the indirect overlap case with fixed occupancies I would imagine that this would be possible. Anyway, just wanted to report on the results of my first post. If anyone knows what I can do to calculate the intensity or simply if it should or shouldn't work (and/or "why") I would greatly appreciate it. Best, Brad Malone UC Berkeley On Sun, Nov 22, 2009 at 11:36 AM, Brad Malone wrote: > Hi, I'm interested in calculating the Raman intensity for a material with a > small indirect band overlap within LDA (~0.3 eV). As has been mentioned many > times on this forum, Raman intensities cannot be calculated for a pure > metal. But should I expect to be able to calculate the intensity for a > material with a small, indirect band overlap? When I attempted to do so > (using lraman=.true. in the ph.x calculation) I received the following > messages: > >Computing Pc [DH,Drho] |psi> >> >> Derivative coefficient: 0.001000Threshold: 1.00E-12 >> ik 1 ibnd 17 linter: root not converged 0.578E+08 >> kpoint 1 ibnd 17 pcgreen: root not converged 0.763E+05 >> ik 1 ibnd 17 linter: root not converged 0.193E+11 >> kpoint 1 ibnd 17 pcgreen: root not converged 0.119E+05 >> ik 1 ibnd 17 linter: root not converged 0.595E+09 >> kpoint 1 ibnd 17 pcgreen: root not converged 0.182E+05 >> ik 1 ibnd 17 linter: root not converged 0.142E+31 >> kpoint 1 ibnd 17 pcgreen: root not converged 0.224E+24 >> ik 1 ibnd 17 linter: root not converged 0.143E+30 >> kpoint 1 ibnd 17 pcgreen: root not converged 0.170E+25 >> ik 1 ibnd 17 linter: root not converged 0.765E+28 >> kpoint 1 ibnd 17 pcgreen: root not converged 0.151E+25 >> ik 1 ibnd 17 linter: root not converged 0.643E+09 >> kpoint 1 ibnd 17 pcgreen: root not converged 0.460E+05 >>
[Pw_forum] problems calculating large numbers of empty states
Hi, I'm trying to calculate a large number of conduction band states for cubic Si (to be used in a subsequent GW calculation). However, when doing this I sometimes run into errors like "problems computing cholesky decomposition" or "too many bands are unconverged". An example file of the nscf calculation that generates an error for me is below: The problems can sometimes be avoided by going to a larger wavefunction cutoff, but I'm not sure if this is for good reason or if changing the wavefunction cutoff just happens to avoid the error in the cases I've tried. Also, for a chosen grid (not the one listed below) I get errors when trying to calculate 400,500, and 700 total bands, but the calculation works fine for 600 bands. Any insight to why this might happen or suggestions as to how to avoid it? Thanks, Brad Malone UC Berkeley - prefix = 'si' calculation = 'nscf' restart_mode = 'from_scratch' wf_collect = .true. tstress = .false. tprnfor = .false. outdir = './' wfcdir = './' pseudo_dir = './' / ibrav = 0 celldm(1) = 10.2612 nat = 2 ntyp = 1 nbnd = 300 ecutwfc = 40.0 / conv_thr = 1.0d-10 diagonalization = 'cg' diago_full_acc = .true. / CELL_PARAMETERS cubic +0.0 +0.5 +0.5 +0.5 +0.0 +0.5 +0.5 +0.5 +0.0 ATOMIC_SPECIES Si 28.086 Si.UPF ATOMIC_POSITIONS crystal Si -0.12500 -0.12500 -0.12500 Si +0.12500 +0.12500 +0.12500 K_POINTS crystal 8 0.0 0.0 0.0 1.0 0.0 0.0 0.25000 8.0 0.0 0.0 0.5 4.0 0.0 0.25000 0.25000 6.0 0.0 0.25000 0.5 24.0 0.0 0.25000 0.75000 12.0 0.0 0.5 0.5 3.0 0.25000 0.5 0.75000 6.0 -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20091022/92715f32/attachment.htm
[Pw_forum] Tips for vc-relax when far from equilibrium
Hi, I am trying to relax (via vc-relax) structures that are, at least most likely, very far from equilibrium. I was wondering if there were any tips on how to make this more robust, as I'm currently experiencing frequent crashes for these systems. The main problem I'm having is shown below: After a *couple successful relaxation steps*, I'll find the following extrapolated charge6.83616, renormalised to8.0 > > total cpu time spent up to now is614.83 secs > > per-process dynamical memory:28.6 Mb > > Self-consistent Calculation > > iteration # 1 ecut=20.00 Ry beta=0.10 > CG style diagonalization > WARNING pzsteqr, convergence not achieved INFO =3 > c_bands: 4 eigenvalues not converged > WARNING pzsteqr, convergence not achieved INFO =3 > c_bands: 4 eigenvalues not converged > WARNING pzsteqr, convergence not achieved INFO =3 > c_bands: 4 eigenvalues not converged > c_bands: 1 eigenvalues not converged > 4 processes killed (possibly by Open MPI) > So the first thing is that my extrapolated charge is very far off from the value of 8 expected (I have two Si atoms in the unit cell). Secondly, the self-consistent calculation fails and kills the job. I've tried using davidson diagonalization as well and it also crashes. Any suggestions on how I might avoid this? Thanks for your time. Best, Brad UC Berkeley -- next part -- An HTML attachment was scrubbed... URL: http://www.democritos.it/pipermail/pw_forum/attachments/20091006/064b83d4/attachment.htm