2014-04-08 19:44 GMT+00:00 Chengbo Zhu <cz...@uowmail.edu.au>: > > Thank you all for your replies. > > Another question: > > 1. if the SR is relaxed in SIESTA, is it necessary to relax it again >> in TranSIESTA when calculating the transmission. (I ran several testing >> model, the SR is optimized again in TranSIESTA) >> > No typically this is not necessary. If your forces on the SR structure for > V=0 are huge, then you might consider relaxing the structure with > transiesta. However I would suspect that something else is wrong. It could > mean that your SR region is too small. > > > In the case I want to see the influence of bias voltage (when V is not > equal to 0) on the structure, I have to relax the structure in transiesta, > is this correct? > Sorry, that was probably not clear from my last mail. No, you typically will not need to relax it after applying a bias. As suggested in the manual, do one siesta relaxation. Then analyze that structure with transiesta, both @V=0 and @|V|>0.
Note that this is the general way of performing the transiesta calculations, and merely a guidance. There could be other things to worry about if you are relaxing the structure, as mentioned in the previous mail (and non-mentioned things!). > > > Regards, > Chengbo > > > > > > > > On 9 Apr 2014, at 0:52, Nick Papior Andersen <nickpap...@gmail.com> > wrote: > > > > > 2014-04-08 13:24 GMT+00:00 Chengbo Zhu <cz...@uowmail.edu.au>: > >> >> Dear SIESTA users, >> >> I am a little confused using TranSIESTA. >> >> It is advised that the geometry of scattering region(SR) should be >> optimised in SIESTA before calculating its transmission in TranSIESTA. >> >> My questions are: >> >> 1. if the SR is relaxed in SIESTA, is it necessary to relax it again in >> TranSIESTA when calculating the transmission. (I ran several testing model, >> the SR is optimized again in TranSIESTA) >> > No typically this is not necessary. If your forces on the SR structure for > V=0 are huge, then you might consider relaxing the structure with > transiesta. However I would suspect that something else is wrong. It could > mean that your SR region is too small. > > >> 2. can I relax the SR in TranSIESTA directly. (if MD.NumCGsteps is set to >> a non-zero number in Transiesta) >> > I have never tested this in the older versions. But I would suspect that > the forces quickly go wild if you don't set this option: > TS.UpdateDMCROnly F > (it then updates the forces on the electrode-connecting atoms) > > I have been told that older versions had difficulties in retaining > charges in geometry optimizations/MD, which invalidates the forces. Hence, > you should be sure that the mulliken charges are "correct". :) > > >> 3. what is the difference between 1 and 2. In a TranSIESTA run, the >> Hamiltonian matrix of SR (and atomic force) is calculated first using >> SIESTA, then TranSIESTA begins by reading two TSHS files. So what is the >> point to run optimization of SR in SIESTA separately from TranSIESTA. >> > You should relax with siesta, as soon as you start transiesta the fermi > level is fixed. > If you start with a far from equilibrium structure you will not relax the > fermi level to the relaxed system which could pose some unforeseen > problems. Again, if your forces in SR are huge (for V=0) compared to > siesta, then something else could be wrong. > >> >> Any thoughts would be appreciated. >> Thank you in advanced. > > >> Regards, >> Chengbo > > > > > -- > Kind regards Nick > > > -- Kind regards Nick
