[Wien] Emax in case.in1 within spin-orbit coupling calculations
Dear all, I have some question on calculation including spin-orbit coupling (SOC). In according to the manual, the second-variational procedure requires to include many more unoccupied states to be calculated. And it can be controlled by increasing the energy maximum (Emax) which eigenvalues shall be searched. For the Emax which defines the number of eigenvalues to be calculated, I have used Emax=2.5 Ry in non-SOC and 5.0 Ry for SOC. Thus, the number of eigenvalues in SOC calculation were increased by about 2 times larger than that of non-SOC. And also I have check the convergence in the function of Emax value. For non-SOC calculations, it does not depend on Emax value as expected. However in the case of SOC calculations, total energy was depend on the Emax value. Larger Emax results the total energy go down. But up to 10Ry of Emax, I cannot find convergence in total energy. You can find the numbers below. system : Bismuth bulk (including SOC) ; total energy Emax = 2 ; E(total) = -86326.2333 Ry Emax = 2.5 ; E(total) = -86326.2420 Ry Emax = 3 ; E(total) = -86326.2454 Ry Emax = 4 ; E(total) = -86326.2527 Ry Emax = 5 ; E(total) = -86326.2577 Ry Emax = 7 ; E(total) = -86326.2634 Ry Emax = 10 ; E(total) = -86326.2662 Ry Questions are, 1. Why the unoccupied states affect the total energy in the case of SOC been included? 2. Is there appropriate suggestions taking Emax value? 3. Should the total energy converged as the function of Emax? 4. Within same Emax restriction (i.e. same parameters), could the total energy be compared between different geometries? Thank you. Best regards, Hyun-Jung Kim. Hyun-Jung Kim (Ph.D student)| phone : ++82 10 7335 7889 Department of Physics | Hanyang University | e-mail: angpangmokjang at hanmail.net 17 Haengdang-Dong | 133-791 Seongdong-Ku,Seoul/Korea| www: http://physics.hanyang.ac.kr/~sst/ -- next part -- An HTML attachment was scrubbed... URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20120319/ae2cbfec/attachment.htm
[Wien] Emax in case.in1 within spin-orbit coupling calculations
Dear all, I have some question on calculation including spin-orbit coupling (SOC). In according to the manual, the second-variational procedure requires to include many more unoccupied states to be calculated. And it can be controlled by increasing the energy maximum (Emax) which eigenvalues shall be searched. For the Emax which defines the number of eigenvalues to be calculated, I have used Emax=2.5 Ry in non-SOC and 5.0 Ry for SOC. Thus, the number of eigenvalues in SOC calculation were increased by about 2 times larger than that of non-SOC. And also I have check the convergence in the function of Emax value. For non-SOC calculations, it does not depend on Emax value as expected. However in the case of SOC calculations, total energy was depend on the Emax value. Larger Emax results the total energy go down. But up to 10Ry of Emax, I cannot find convergence in total energy. You can find the numbers below. system : Bismuth bulk (including SOC) ; total energy Emax = 2 ; E(total) = -86326.2333 Ry Emax = 2.5 ; E(total) = -86326.2420 Ry Emax = 3 ; E(total) = -86326.2454 Ry Emax = 4 ; E(total) = -86326.2527 Ry Emax = 5 ; E(total) = -86326.2577 Ry Emax = 7 ; E(total) = -86326.2634 Ry Emax = 10 ; E(total) = -86326.2662 Ry Questions are, 1. Why the unoccupied states affect the total energy in the case of SOC been included? 2. Is there appropriate suggestions taking Emax value? 3. Should the total energy converged as the function of Emax? 4. Within same Emax restriction (i.e. same parameters), could the total energy be compared between different geometries? Thank you. Best regards, Hyun-Jung Kim. Hyun-Jung Kim (Ph.D student)| phone : ++82 10 7335 7889 Department of Physics | Hanyang University | e-mail: angpangmokjang at hanmail.net 17 Haengdang-Dong | 133-791 Seongdong-Ku,Seoul/Korea| www: http://physics.hanyang.ac.kr/~sst/ -- next part -- An HTML attachment was scrubbed... URL: http://zeus.theochem.tuwien.ac.at/pipermail/wien/attachments/20120319/be9a99a3/attachment.htm
[Wien] Emax in case.in1 within spin-orbit coupling calculations
Just as the total energy will decrease asymptotically (=forever) if you increase RKMax, it will decrease asymptotically if you increase Emax. How do you find the appropriate RKMax? == not by monitoring the total energy as such, but either energy *differences* or your observable property of interest. The same applies here: either you observe the energy difference (between e.g. two cases with a different magnetic order), or you monitor an observable property that depends on spin-orbit coupling (e.g. orbital contribution to the magnetic moment). Stefaan On 19/03/2012 11:26, hyunjung kim wrote: Dear all, I have some question on calculation including spin-orbit coupling (SOC). In according to the manual, the second-variational procedure requires to include many more unoccupied states to be calculated. And it can be controlled by increasing the energy maximum (Emax) which eigenvalues shall be searched. For the Emax which defines the number of eigenvalues to be calculated, I have used Emax=2.5 Ry in non-SOC and 5.0 Ry for SOC. Thus, the number of eigenvalues in SOC calculation were increased by about 2 times larger than that of non-SOC. And also I have check the convergence in the function of Emax value. For non-SOC calculations, it does not depend on Emax value as expected. However in the case of SOC calculations, total energy was depend on the Emax value. Larger Emax results the total energy go down. But up to 10Ry of Emax, I cannot find convergence in total energy. You can find the numbers below. system : Bismuth bulk (including SOC) ; total energy Emax = 2 ; E(total) = -86326.2333 Ry Emax = 2.5 ; E(total) = -86326.2420 Ry Emax = 3 ; E(total) = -86326.2454 Ry Emax = 4 ; E(total) = -86326.2527 Ry Emax = 5 ; E(total) = -86326.2577 Ry Emax = 7 ; E(total) = -86326.2634 Ry Emax = 10 ; E(total) = -86326.2662 Ry Questions are, 1. Why the unoccupied states affect the total energy in the case of SOC been included? 2. Is there appropriate suggestions taking Emax value? 3. Should the total energy converged as the function of Emax? 4. Within same Emax restriction (i.e. same parameters), could the total energy be compared between different geometries? Thank you. Best regards, Hyun-Jung Kim. Hyun-Jung Kim (Ph.D student)| phone : ++82 10 7335 7889 Department of Physics| Hanyang University| e-mail: angpangmokjang at h mailto:hyunjung at fhi-berlin.mpg.deanmail.net http://anmail.net 17 Haengdang-Dong| 133-791 Seongdong-Ku,Seoul/Korea| www: http://physics.hanyang.ac.kr/~sst/ ___ Wien mailing list Wien at zeus.theochem.tuwien.ac.at http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
