Hi, Den fre. 10. jul. 2020 kl. 05.00 skrev El-abed Haidar < ehai2...@uni.sydney.edu.au>:
> Thank you a lot for the outline Nick and thank you Berna for your insight. > > > > So coming back to my email that means it is feasible to start with lower K > points (as long as they are converged) and then increase them only in > Tbtrans. > > A couple of questions though: > > 1. @Nick Papior <nickpap...@gmail.com> when you increase your k points > in Tbtrans that wont change the kpoints in TSHS files because those are > TranSIESTA. Is that correct? Because that affects the Eigenchannel > calculations. > > No, the TSHS is the output of TranSiesta. TSHS files are only read by TBtrans, not altered or changed. If you mean EigenChannels from the Inelastica suite, then the k-points in your tbtrans calculation has no influence on the Inelastica calculation, I am not sure exactly what you mean here... > > 1. I am asking this because when I read any paper based on TranSIESTA > and they mention k points 1 1 100 i would assume that means in TbTrans they > have done the same. > > If the paper you read were mentioning this for the electrode region then it suggests to me that the system is a "chain" and thus have no k-points transverse to the transport direction which means that you don't need to use k-points in TBtrans. So this is perfectly fine. I would advice to always disclose 3 k-point samplings in any publication, 1) the electrode k-point sampling, 2) the transiesta k-point sampling, 3) the tbtrans k-point sampling. If you are simulating chains, then you don't only need to mention the electrode, for obvious reasons :) > > 1. So in summary we should do a SIESTA convergence test and then > Tbtrans. Would you agree? > > Yes. And if you want to really be sure, then once you have converged TBtrans, you do a SIESTA with increased k-points, then redo the tbtrans for the converged tbtrans k-point sampling. This will really show the robustness of the parameters. :) > Thank you all ! > > EL-abed > > El-abed Haidar | Doctor of Philosophy (Science) > Condensed Matter Theory (CMT) Group| School of Physics > THE UNIVERSITY OF SYDNEY | NSW | 2006 > > > > > > *From: *Nick Papior <nickpap...@gmail.com> > *Sent: *Friday, 10 July 2020 6:04 AM > *To: *siesta-l <siesta-l@uam.es> > *Subject: *Re: [SIESTA-L] K points in TranSIESTA vs TbTrans question > > > > The basic principle of transiesta is as with siesta. > > > > A rough outline would be this: > > > > 0) Create your system, decide upon the electrodes. Ensure your device > screens off the defect towards the electrodes (meaning you'll need some > additional electrode layers) > > 1) You converge your density in your electrodes. This is a standard > procedure for *any* DFT calculations, check all parameters k-points, > mesh-cutoff, pseudos etc. > > 2) Once you have defined your converged parameters, you try those on the > full device system, here starting with a regular siesta calculation should > suffice. Once done you should again ensure that you converge the system > parameters, i.e. try increasing k-points etc and see if you get the same > results. > > 3) Now you do your transiesta calculation with your converged parameters. > > 4) TBtrans calculates DOS and transmissions. These quantities are more > sensitive to the k-point sampling compared to the charge density. So you, > generally, should always use a denser k--grid for tbtrans. > > See slide 5 here: > https://github.com/zerothi/ts-tbt-sisl-tutorial/releases/download/v2018.11/talk_2.pdf > <https://protect-au.mimecast.com/s/O7gOCxngwOfMy80Kt826WB?domain=github.com> > > Needless to say, this is also important for correct DOS analysis for > siesta calculations. > > So again you have to converge the number of k-points for your tbtrans > calculations. > > > > Den ons. 8. jul. 2020 kl. 22.00 skrev El-abed Haidar < > ehai2...@uni.sydney.edu.au>: > > Good afternoon, > > I was wondering about the following: > > 1. I have done a TranSIESTA + TbTrans calculation using 1 1 100 and > had a look at transmission curves and PDOS. I got certain results but I > then increased k points to 5 1 100. And only ran a TbTrans. The NEW > results saw some changes. > 2. I decided then to restart from scratch the same Transiesta + > Tbtrans calculation with 5 1 100 instead of 1 1 100. I got almost the same > results as the NEW in step 1. Is that expected? If so why ? > 3. If so also does that mean I could start with less k points and just > increase them in the TbTrans run by using TBT.k [a b c]? > > > > I just want to make sure that my results are feasible and acceptable. > > Thank you and looking forward to your reply. > > EL-abed > > El-abed Haidar | Doctor of Philosophy (Science) > Condensed Matter Theory (CMT) Group| School of Physics > THE UNIVERSITY OF SYDNEY | NSW | 2006 > > > > > > > -- > SIESTA is supported by the Spanish Research Agency (AEI) and by the > European H2020 MaX Centre of Excellence (http://www.max-centre.eu/ > <https://protect-au.mimecast.com/s/c_EqCyojxQTXgDWEsMKUIG?domain=max-centre.eu/> > ) > > > > > -- > > Kind regards Nick > > > -- Kind regards Nick
-- SIESTA is supported by the Spanish Research Agency (AEI) and by the European H2020 MaX Centre of Excellence (http://www.max-centre.eu/)
