Top Page > Browsing

O(N) Calculations for Cases with Spin Orbit Coupling Date: 2015/06/17 02:52 Name: KZ Hi, Dear developers and users, I am using OpenMX to calculate transmission properties for some large systems. Thus I am looking into the O(N) method implemented in OpenMX. And I am also working on cases that have Spin Orbit Coupling effects. But it seems O(N) methods do not apply to Non-Collinear case, right? So I would like to ask: for bands calculation with SOC and transmission with SOC of some large systems, what should be the optimized method in terms of time and accuracy to do the calculation etc.? Besides the parallelization, can I use O(N) method to optimize the calculation more? Thanks. K.Z.

Page: [1]

Re: O(N) Calculations for Cases with Spin Orbit Coupling ( No.1 ) Date: 2015/06/17 18:45 Name: Artem Pulkin If you cannot afford DFT use model Hamiltonians: several have been reported for transition metal dichalcogenides. You can also construct your own using wannierization procedure either in OpenMX or in some other codes. The DFTB is also an option, see this one for example: http://ieeexplore.ieee.org/xpl/articleDetails.jsp?arnumber=6932431 Ofcourse, nothing comes for free: I calculated some of line defects from the above paper with OpenMX and obtained transmission differing by several orders of magnitude. Use with care. Artem Re: O(N) Calculations for Cases with Spin Orbit Coupling ( No.2 ) Date: 2015/06/19 06:11 Name: KZ Hi, Artem, Thanks for the comments. However, I am kind like curious about why Krylov subspace methods do not support non-collinear methods in OpenMX. Could you explain the reason of that? And could I still use Divide-Conquer method to calculate non-collinear cases? It will help me understand and use Order N methods in OpenMX better. Regards, K.Z. Re: O(N) Calculations for Cases with Spin Orbit Coupling ( No.3 ) Date: 2015/06/19 17:30 Name: Artem Pulkin I have never used O(N) in my research. I do not know why spin-orbit case is not implemented (probably because the difference between O(N) and conventional band structure is larger than spin-orbit effects). So wait for someone else to reply. Artem Re: O(N) Calculations for Cases with Spin Orbit Coupling ( No.4 ) Date: 2015/06/24 00:54 Name: T. Ozaki Hi, The O(N) methods are not supported for the non-collinear case, which is due to a rather historical reason. At that time we have been developing the O(N) methods, we were interested in large and complicated structures in which SOI may not play an important role, and thereby we simply ignored the extension to the non-collinear case. If your systems dose not require SOI for the structural optimization, I would recommend you to perform the O(N) calculation for structural optimization, and after getting the optimized structure, you can perform the conventional diagonalization for the optimized structure obtained from the first calculation. If SOI plays an important role even for the geometrical structure such as surface structure of Bi, we need to perform the non-collinear calculation from the beginning. Regards, TO Re: O(N) Calculations for Cases with Spin Orbit Coupling ( No.5 ) Date: 2015/06/24 10:33 Name: KZ Hi, Dr. Ozaki, Since I am using OpenMX to perform NEGF calculation for large systems. And I want to include SOC for the transmission calculations. Could you help suggest which way of calculation for SOC of OpenMX should be better? Could I use the O(N) methods in OpenMX? Regards, K.Z

Page: [1]