[Wien] A question regarding a calculation for a 56 atoms unit cell
Are you using a recent version with MSEC1 in case.inm or an older one ? As was mentioned several times gefore: changing the mixing parameter will most likely NOT help with the new methods. Using TEMP instead of TETRA may help convergence (but a too large value is dangerous). Do you have a metallic or insulating state (I understand it SHOULD be an insulator, but do you get it ?) How many iterations did you run ? Sometimes (but not very often) more than 40 are required .. Timing is strange. 56 atoms is not that big and it should not take a week ... (? 2-fold k-point parallelization ?; OMP_NUM_THREAD =2??? to utilize all 4 cores ??) I guess nobody can answer if your convergence is sufficient or not. It depends on the prupose of your calculation. You need to monitor other quantities like MMT or MMIxxx, FER, forces, :GAP, Am 05.06.2011 22:17, schrieb shamik chakrabarti: Dear wien2k users, We are running a SCF calculation for a *56 inequivalent atoms* unit cell. We have included spin polarization and self interaction correction (GGA+U) in this calculation. The calculation has been run for one week now without any errors. We are using a system having *Intel Xeon quad core processor and 8GB RAM*. We are using *14 K points* and a *mixing factor 0.1*. The energy and charge convergence were achieved up to *0.04*. Once energy convergence has also gone up to 0.008 but then again a bit diverges to come up to 0.04. Now my question is:: (1) Is it possible to achieve charge and energy convergence up to 0.0001 with this unit cell using the above mentioned system if we consider that all the interactions applied for this material is proper? (2) If it is achievable then can we estimate how much time it can take given the no. of inequivalent atoms, K points, mixing parameter and nature of the processor? (3) If we are not able to achieve that much of convergence (0.0001) even after a long time (say 2 weeks) then is it right to stop the calculation if we achieve a stability in convergence and say it is up to 0.01? (4) If we take the convergence up to 0.01 then how much reliable will be the further calculation of its properties (DOS) given the no. of K points is 14? (5) If we do not get the convergence say at least up to 0.01 is it right to increase the mixing parameter to say 0.15 or 0.20 given the system is a *localized magnetic system*? The system is an insulator and also contain Fe. Thats why we have also consider self interaction correction (GGA+U). Please forgive us for such a long email. We have to right it in this way as the calculation has already been taken a long time and are indicating to take longer time still. Any reply in this regard will be very helpful for us. Thanks in advance. with best regards, -- Shamik Chakrabarti Research Scholar Dept. of Physics Meteorology Material Processing Solid State Ionics Lab IIT Kharagpur Kharagpur 721302 INDIA ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien -- P.Blaha -- Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna Phone: +43-1-58801-15671 FAX: +43-1-58801-15698 Email: blaha at theochem.tuwien.ac.atWWW: http://info.tuwien.ac.at/theochem/ --
[Wien] A question regarding a calculation for a 56 atoms unit cell
Dear Dr. Peter Blaha Sir, We are using a version with MSEC1 in case.inm Sir it was experimentally verified that it is an insulator (conductivity in the range of 10^ -7 in room temperature). Each iteration is taking 1hr. 50 mins and it has crossed already more than 60 iterations (considering time loss of around 10 hr due to power failure*but each time we have started from where it ends...we have used run anyway option)* * * Sir we have set OMP_NUM_THREAD = 8 and it using all the 8 processors as can also been seen from CPU_MONITOR Sir we will wait for another week and see what happens. We have data for experimental band gap and we can compare it with the theoretically obtained one. Sir thank you very much for your reply. with best regards, Shamik Chakrabarti On Mon, Jun 6, 2011 at 11:47 AM, Peter Blaha pblaha at theochem.tuwien.ac.atwrote: Are you using a recent version with MSEC1 in case.inm or an older one ? As was mentioned several times gefore: changing the mixing parameter will most likely NOT help with the new methods. Using TEMP instead of TETRA may help convergence (but a too large value is dangerous). Do you have a metallic or insulating state (I understand it SHOULD be an insulator, but do you get it ?) How many iterations did you run ? Sometimes (but not very often) more than 40 are required .. Timing is strange. 56 atoms is not that big and it should not take a week ... (? 2-fold k-point parallelization ?; OMP_NUM_THREAD =2??? to utilize all 4 cores ??) I guess nobody can answer if your convergence is sufficient or not. It depends on the prupose of your calculation. You need to monitor other quantities like MMT or MMIxxx, FER, forces, :GAP, Am 05.06.2011 22:17, schrieb shamik chakrabarti: Dear wien2k users, We are running a SCF calculation for a *56 inequivalent atoms* unit cell. We have included spin polarization and self interaction correction (GGA+U) in this calculation. The calculation has been run for one week now without any errors. We are using a system having *Intel Xeon quad core processor and 8GB RAM*. We are using *14 K points* and a *mixing factor 0.1*. The energy and charge convergence were achieved up to *0.04*. Once energy convergence has also gone up to 0.008 but then again a bit diverges to come up to 0.04. Now my question is:: (1) Is it possible to achieve charge and energy convergence up to 0.0001 with this unit cell using the above mentioned system if we consider that all the interactions applied for this material is proper? (2) If it is achievable then can we estimate how much time it can take given the no. of inequivalent atoms, K points, mixing parameter and nature of the processor? (3) If we are not able to achieve that much of convergence (0.0001) even after a long time (say 2 weeks) then is it right to stop the calculation if we achieve a stability in convergence and say it is up to 0.01? (4) If we take the convergence up to 0.01 then how much reliable will be the further calculation of its properties (DOS) given the no. of K points is 14? (5) If we do not get the convergence say at least up to 0.01 is it right to increase the mixing parameter to say 0.15 or 0.20 given the system is a *localized magnetic system*? The system is an insulator and also contain Fe. Thats why we have also consider self interaction correction (GGA+U). Please forgive us for such a long email. We have to right it in this way as the calculation has already been taken a long time and are indicating to take longer time still. Any reply in this regard will be very helpful for us. Thanks in advance. with best regards, -- Shamik Chakrabarti Research Scholar Dept. of Physics Meteorology Material Processing Solid State Ionics Lab IIT Kharagpur Kharagpur 721302 INDIA ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien -- P.Blaha -- Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna Phone: +43-1-58801-15671 FAX: +43-1-58801-15698 Email: blaha at theochem.tuwien.ac.atWWW: http://info.tuwien.ac.at/theochem/ -- ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien -- Shamik Chakrabarti Research Scholar Dept. of Physics Meteorology Material Processing Solid State Ionics Lab IIT Kharagpur Kharagpur 721302 INDIA -- next part -- An HTML attachment was scrubbed... URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20110606/2656c0c9/attachment.htm
[Wien] A question regarding a calculation for a 56 atoms unit cell
Sir we have set OMP_NUM_THREAD = 8 and it using all the 8 processors as can also been seen from CPU_MONITOR OMP_NUM_THREAD = 8 does NOT give any speed up !! Did you check your running time with values of OMP_NUM_THREAD =1,2,4 and a k-parallelization ! -- P.Blaha -- Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna Phone: +43-1-58801-15671 FAX: +43-1-58801-15698 Email: blaha at theochem.tuwien.ac.atWWW: http://info.tuwien.ac.at/theochem/ --
[Wien] A question regarding a calculation for a 56 atoms unit cell
Sir, No...we have not checked running time with values of OMP_NUM_THREAD =1,2,4...*is it that we should set it to 4 as we are using quad core or we have to check for which value it will give us the highest speed?*.. with best regards, Shamik Chakrabarti On Mon, Jun 6, 2011 at 1:26 PM, Peter Blaha pblaha at theochem.tuwien.ac.atwrote: Sir we have set OMP_NUM_THREAD = 8 and it using all the 8 processors as can also been seen from CPU_MONITOR OMP_NUM_THREAD = 8 does NOT give any speed up !! Did you check your running time with values of OMP_NUM_THREAD =1,2,4 and a k-parallelization ! -- P.Blaha -- Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna Phone: +43-1-58801-15671 FAX: +43-1-58801-15698 Email: blaha at theochem.tuwien.ac.atWWW: http://info.tuwien.ac.at/theochem/ -- ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien -- Shamik Chakrabarti Research Scholar Dept. of Physics Meteorology Material Processing Solid State Ionics Lab IIT Kharagpur Kharagpur 721302 INDIA -- next part -- An HTML attachment was scrubbed... URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20110606/a073ac85/attachment.htm