I have found some discussions suggesting real space approach instead of reciprocal space for large systems. What is the logic behind this real space approach? Dosen't DFT approach always only includes the latter part (like the K points etc)? I have now approach to systems with large number of atoms and I really need to grasp this idea of transition from reciprocal to direct spaces.
The reason is that it is more convenient and faster to calculate some contributions of the total energy in reciprocal space and other ones in real space. For example, since the exchange-correlation energy is local in real space, its calculation is performed on a real space grid. FFT and inverse FFT are used to connect direct and reciprocal spaces within a code.
The use of k-points is most convenient when dealing with periodic system. For aperiodic systems one needs to include a certain amount of vacuum. Therefore at a certain point such vacuum begins to cost significantly more than the actual material. And one might decide to move to a real space description where such a problem does not exist.
Thanks to both Roberto D'Agosta and Rodolphe Pollet for your kind reply and information. Could you also suggest me some basic or fundamental articles/papers so that I might have the idea on performing such tasks regarding the same.
Like for example it is known that for larger systems, the k points can be reduced because larger systems in direct real spaces actually corresponds to smaller reciprocal geometry thus requiring lesser k points. Can this logic still be applied since now one have approached towards real space instead or reciprocal spaces?
You should try to be more specific on what you mean by large system and the type of calculations you want to perform.
It is true that by increasing the unit cell, you can reduce the number of k points as the accuracy remains the same. At the very large unit cell you can reduce to a single k-point, gamma.
DFT per se, is valid for an electron Hamiltonian. How you decide to represent this operator (plane waves, Gaussian, real space) does not change the essence of the theory and is often a matter of taste and numerical choice. If you work with solids, k-space is the natural choice, if you work with molecules, you better work with localised basis sets or real space.
I don't have a the moment a proper reference for discussing this.
Dear Roberto D'Agosta I am working on bulk crystals and currently on doping for their electronic properties. For such simulation it is required to have larger supercells and atoms in my structure now increases to approximately 64 atoms; which is way beyond the capacity of my computational system if real space approach is not implemented.
I strongly recommend this book:
Marx D, Hutter J (2009) Ab initio molecular dynamics: basic theory and advanced methods, Oxford University Press, Oxford.
For DFT calculations, the real quantity is the number of electrons or the number of Kohn-Sham states. Normally you can try to reduce the number of states by controlling the number of hands. Then reduce the k-point sampling. Finally, you can try to look for pseudopotentials without core electrons. But this last step will definitely reduce the accuracy.
Thanks to both Roberto D'Agosta and Rodolphe Pollet for sharing critical information. I will proceed my calculations keeping in mind your advices; but I am sure to face problem as this will be one of a kind approach and I still might need your cooperations in the future.
- Badges
- Science topic
- Similar topics
- Engineering
- Electrical Engineering
- Electrodes