Re: [Wien] SCF convergence
Reduce RKmax. It is ridiculous to start such a calculation with RKmax=9. Probably convergence was spoiled by the occurrence of ghostbands. To do it right: Do a simple wurzite structure. Start with RKmax=5.5 (or even 5.0 ) (see http://www.wien2k.at/reg_user/faq/rkmax.html) and do: run_lapw -fc 1 -cc 0.0001 check gap and forces. save wc_rkm5.5 increase rkmax to 6.5 in case.in1c run_lapw -fc 1 -cc 0.0001check again gap and forces any significant changes ? If not, you have found a reasonable rkmax (5.5) to go to the large calculation. For a BIG calculation ALWAYS start with low convergence parameters (which, if possible, should be tested before on a small system). Once the BIG calculation has converged and is fully relaxed, you can still check your results by increasing RKmax and continuing, but NEVER start with such huge values. On 09/25/2013 04:22 PM, Luis Ogando wrote: Dear Wien2k community, We are trying do calculate the influence of zinc blend (ZB) stacking faults on InP (sp semiconductor) wurtzite (WZ) systems using Wien2k 13. Our first goal is to calculate the band gap change along the c-axis (perpendicular to the interface between the two phases), similar to what was done in J. Appl. Phys. 114, 033709 (2013) for two semi-infinite structures. In our case, we use a supercell with three ZB cells along (111) (hexagonal form) matched along c to a large sequence of WZ cells to guarantee a minimum reproduction of a ZB environment in the stacking fault region. We have done calculations for a system with 10 WZ cells, but the gap did not change along the structure and we are now trying 15 WZ cells and the 3 ZB ones (78 atoms in total) to isolate neighbour stacking faults. As the gap is our main property, we are using the mBJ potential, but we are facing convergence problems at the first (PBE) SCF calculation. We would be glad if someone could give us any hint about how to improve the PBE scf convergence. Here we give some relevant parameters of our calculation (please, do not hesitate in asking others): *Species: In and P *Number of atoms: 78 *RMT: 2.50 (In) and 2.10 (P) (NN-DIST= 4.84768) *RKmax: 9 *Emax: 5 (spin-orbit effects) *XC: 13 (PBE) *core-valence separation energy: -6.0 *GMAX: 20 *Inequivalent Kpoints: 12 increased to 24 after some iterations *Mixing schema: MSR1 changed to PRATT with 0.15 mixing factor after some iterations *Calculation options: run_lapw -p -NI -ec 0.0001 -cc 0.0001 -i 150 -it ** We also attached the energy (Energy.dat) and charge (Charge.dat) convergence evolution. Many thanks in advance. All the best, Luis ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html -- P.Blaha -- Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna Phone: +43-1-58801-165300 FAX: +43-1-58801-165982 Email: bl...@theochem.tuwien.ac.atWWW: http://info.tuwien.ac.at/theochem/ -- ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
[Wien] re-initialization issue with mBJ on reducing RMT
Dear Wien 2k users, I am working with Wien2k 12.1 for calculating electronic and optical properties of MoTeSe using a 1x1x2 supercell. Scf convergence with pbe-sol gga was successful and error free (I reduced the RMT by 5%) so I proceeded with mBJ calculations. While running the scf with mBJ potential, I got an error in LAPW1: Select: No energy limit found for atom2, L=0 Select: Ebottom 2.97919, Etop -200.000. Now I am trying to reduce the RMT more or change the energy between core and valence states to rectify this error. If I do that, do I need to reinitialize and run pbe sol again and then switch to mBJ calc? or can I directly re-initialize in some way for mBJ calc only? rgds Ushma Ahuja ML Sukhadia University Udaipur, India___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] re-initialization issue with mBJ on reducing RMT
Hi, Yes, it is better to reinitialize and run PBEsol again before switching to mBJ. F. Tran On Thu, 26 Sep 2013, Ushma Ahuja wrote: Dear Wien 2k users, I am working with Wien2k 12.1 for calculating electronic and optical properties of MoTeSe using a 1x1x2 supercell. Scf convergence with pbe-sol gga was successful and error free (I reduced the RMT by 5%) so I proceeded with mBJ calculations. While running the scf with mBJ potential, I got an error in LAPW1: Select: No energy limit found for atom2, L=0 Select: Ebottom 2.97919, Etop -200.000. Now I am trying to reduce the RMT more or change the energy between core and valence states to rectify this error. If I do that, do I need to reinitialize and run pbe sol again and then switch to mBJ calc? or can I directly re-initialize in some way for mBJ calc only? rgds Ushma Ahuja ML Sukhadia University Udaipur, India ___ Wien mailing list Wien@zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien SEARCH the MAILING-LIST at: http://www.mail-archive.com/wien@zeus.theochem.tuwien.ac.at/index.html
Re: [Wien] SCF convergence
Dear Prof. Blaha, Thank you for your comments. I have reasons to change the default parameters and I would like to clarify some points: 1) All the structural parameters were taken from optimized WZ and ZB bulk InP structures using the PBE potential. 2) The increase in the number o K-points I mentioned in my first message was done only in the a-b (x-y) plane, because there is no sense in increasing K-points along the huge c-axis of the supercell. 3) Some time ago, there was a message sent to the list saying that an increase in the Gmax value could improve the calculation performance. I did some tests in the simple WZ structure and I got a faster calculation with Gmax=20. This was (more or less) the value cited in the message, but, unfortunately, I do not remember details (the message is somewhere in my mailbox). 4) As you noted, the gap is the critical quantity in my calculations and I did the suggested RKmax tests some time ago. The main parameters in these tests were (only Gmax and Kpoints are different from that employed in the supercell): *Structure: WZ InP bulk *RMT: 2.50 (In) and 2.10 (P) *Emax: 5 (spin-orbit effects) *XC: 13 (PBE) + mBJ *core-valence separation energy: -6.0 *GMAX: 12 (NOT 20 !!!) *Inequivalent Kpoints: 16 *Calculation options: run_lapw -p -NI -so -ec 0.0001 -cc 0.0001 -i 150 -it Results: RKmax = 5 = gap = 1.556 eV RKmax = 6 = gap = 1.467 eV RKmax = 7 = gap = 1.476 eV RKmax = 8 = gap = 1.486 eV RKmax = 9 = gap = 1.493 eV ( the experimental value is 1.49 eV ; Appl. Phys. Lett. 91, 263104 (2007) ) RKmax = 10 = gap = 1.498 eV The difference between the calculated gaps for WZ and ZB bulk structures is around 98 meV, but I doubt that the 3 ZB layers will reproduce the bulk gap, so the difference between phases in the supercell must be lower than 98 meV. This is the reason for using RKmax = 9, I would like to get the best gap value for the WZ region. In your message, you commented that the convergence may be disturbed by ghostbands. Do you believe that the -in1new option can help ? I would strongly appreciate any comment / suggestion ! Thank you again, Luis 2013/9/26 Peter Blaha pbl...@theochem.tuwien.ac.at Reduce RKmax. It is ridiculous to start such a calculation with RKmax=9. Probably convergence was spoiled by the occurrence of ghostbands. To do it right: Do a simple wurzite structure. Start with RKmax=5.5 (or even 5.0 ) (see http://www.wien2k.at/reg_user/** faq/rkmax.html http://www.wien2k.at/reg_user/faq/rkmax.html) and do: run_lapw -fc 1 -cc 0.0001 check gap and forces. save wc_rkm5.5 increase rkmax to 6.5 in case.in1c run_lapw -fc 1 -cc 0.0001check again gap and forces any significant changes ? If not, you have found a reasonable rkmax (5.5) to go to the large calculation. For a BIG calculation ALWAYS start with low convergence parameters (which, if possible, should be tested before on a small system). Once the BIG calculation has converged and is fully relaxed, you can still check your results by increasing RKmax and continuing, but NEVER start with such huge values. On 09/25/2013 04:22 PM, Luis Ogando wrote: Dear Wien2k community, We are trying do calculate the influence of zinc blend (ZB) stacking faults on InP (sp semiconductor) wurtzite (WZ) systems using Wien2k 13. Our first goal is to calculate the band gap change along the c-axis (perpendicular to the interface between the two phases), similar to what was done in J. Appl. Phys. 114, 033709 (2013) for two semi-infinite structures. In our case, we use a supercell with three ZB cells along (111) (hexagonal form) matched along c to a large sequence of WZ cells to guarantee a minimum reproduction of a ZB environment in the stacking fault region. We have done calculations for a system with 10 WZ cells, but the gap did not change along the structure and we are now trying 15 WZ cells and the 3 ZB ones (78 atoms in total) to isolate neighbour stacking faults. As the gap is our main property, we are using the mBJ potential, but we are facing convergence problems at the first (PBE) SCF calculation. We would be glad if someone could give us any hint about how to improve the PBE scf convergence. Here we give some relevant parameters of our calculation (please, do not hesitate in asking others): *Species: In and P *Number of atoms: 78 *RMT: 2.50 (In) and 2.10 (P) (NN-DIST= 4.84768) *RKmax: 9 *Emax: 5 (spin-orbit effects) *XC: 13 (PBE) *core-valence separation energy: -6.0 *GMAX: 20 *Inequivalent Kpoints: 12 increased to 24 after some iterations *Mixing schema: MSR1 changed to PRATT with 0.15 mixing factor after some iterations *Calculation options: run_lapw -p -NI -ec 0.0001 -cc 0.0001 -i 150 -it ** We also attached the energy (Energy.dat) and charge (Charge.dat) convergence evolution. Many thanks in advance. All the best,
