Top Page > Browsing

O(N) method in Openmx Date: 2014/09/01 05:59 Name: Chen Dear Prof. Ozaki, I am very interested in the O(N) method of Openmx. The capability of dealing with more than ten thousand atoms is very impressive. I am now studying a junction structure with plane-wave method but the computational demand is huge (the number of atoms > 5000). I am trying to use the O(N) methods (DC,Krylov,low-order scaling) in Openmx. If I understand correctly, there is a compromise between accuracy and efficiency for O(N) method. For the junction structure, e.g. Fe/GaAs junction, the relative alignment of Fermi levels of Fe and GaAS is quite important for me. If I use the conventional band structure method, it is not difficult to determine the Fermi level since we can fill up the electronic states until the number of electrons is reached. I am wondering if O(N) method can predict the position of the Fermi level of a material (e.g. bulk) accurately. Will the Fermi level change with the choice of input parameters, i.e. the cluster size? If one of your paper has already mentioned this point, please tell me. Thanks a lot. Chen

Page: [1]

Re: O(N) method in Openmx ( No.1 ) Date: 2014/10/27 22:41 Name: T. Ozaki Hi, In order to determine the Fermi level, you can combine both the O(N) and conventional diagonalization methods as explained at http://www.openmx-square.org/openmx_man3.7/node76.html Recent papers on application of the O(N) may give a crue on how the O(N) can be applied for your system: T. Ohwaki et al., J. Chem. Phys. 140, 244105 (2014). T. Ohwaki et al., J. Chem. Phys. 136, 134101 (2012). H. Sawada et al., Modelling Simul. Mater. Sci. Eng. 21, 045012 (2013). N. Ileri et al., Theo. Chem. Accounts 133, 1575 (2014). Best regards, TO

Page: [1]