Relative speed and accuracy of device algorithm, exchange correlation, basis set

[Anirban Basak]

Hi,

        I want to know the relative speed and accuracy of device algorithm (maybe like Recursion > Direct > Krylov), exchange Correlation, basis set for ATK-DFT (device). (especially in the case of I-V characteristics of CNT with various metal contacts). Accuracy is very important.

I'd also like to know whether ATK-device calculates both ballistic transport and diffusive transport. Does not ATK calculate scattering probability from scattering potentials in diffusive transport?

Thank You

[zh]

The available exchange-correlation functional in ATK-DFT includes the LDA and GGA.  Actually, both of them are the same level. It is difficult to give a definitized suggestion to say which one of them is much better for the problem under your study.  The job for the GGA calculation may somehow slightly heavier than that of the LDA, but usually it can be ignored.

Larger basis set, heavier job and more accurate. In other words,  for accuracy: DZP > SZP > SZ; for speed:  SZ> SZP> DZP

Only the ballistic transport can be calculated, the diffusive transport cannot be.

[Anirban Basak]

What about the device algorithms? Can you make a relative comparison between Krylov, Direct and Recursion methods?

[Anders Blom]

The difference is not dramatic, perhaps Krylov is 10% faster, but it depends on the system at hand, what dominates the time consumption. It's not a major point of performance under any circumstance.

[Anirban Basak]

So. No diffusive transport? Only Ballistic Transport.