Peter Elliott

Hi Trishu, I tried your 2 input files and get the same kpoint grid for both. As you say, the kpoint grid shouldn't change if you just double your magnetic field. So that behavior shouldn't happen. I suspect the issue is with the compiler, but I don't have much knowledge about them. Could you give some more details on the version of ELK and the compilers and libraries you used. -Peter

Hi Jerry, ELK finds the Fermi level by finding the value that makes the sum of the occupancies equal to the number of particles. The Fermi level is found by a bisection method. The form of occupancy depends on the smearing method you choose, but the default is just the Fermi-Dirac occupation. The default smearing is equivalent to a temperature of about 300K if I remember correctly. You can decrease this using tempk, although you can't go too low otherwise things probably don't converge. So if you...

just some more notes on this: in tdrestart, potkst needs the atomic positions datposc, which are currently read in after tdrestart gengvclns needs rho static which is read in tdrestart so i'd do readrhos back in tddft instead of tdrestart then in tdinit do all the Ehrenfest stuff before the tdrestart stuff and then just before tdrestart, call readdatposc if tdatpos is true

just some more notes on this: in tdrestart, potkst needs the atomic positions datposc, which are currently read in after tdrestart gengvclns needs rho static which is read in so i'd do readrhos back in tddft instead of tdrestart then in tdinit do all the Ehrenfest stuff before the tdrestart stuff and then just before tdrestart, call readdatposc if tdatpos is true

I think the atomic positions are reset during task 463 instead of reading in the current positions. Looking into the files: In tdinit.f90 , the atomic position are read if trddatpos is true In tdrestart.f90, trddatpos is set to true if tdatpos is true However in tdinit.f90 , tdatpos is set to true for task 463 on line 94 after tdrestart has been called on line 87. So I think trddatpos never gets set to true for task 463. -Peter

A niche issue but when running task 320 with a scissor correction and with a non-ground-state set of occupations, then: In genvchi0, if the conduction band has occupation, then transitions within the conduction bands will erroneously get a scissor correction. This can lead to a spurious peak around the scissor correction energy. In task 121, the dielectric routine doesn't have this problem as it checks whether the transition goes from valence to conduction. Copying this code to genvchi0 fixes the...

A niche issue but when running task 320 with a scissor correction and with a non-ground-state set of occupations, then: In genvchi0, if the conduction band has occupation, then transitions within the conduction bands will erroneously get a scissor correction. This can lead to a spurious peak at the scissor correction energy. In task 121, the dielectric routine doesn't have this problem as it checks whether the transition goes from valence to conduction. Copying this code to genvchi0 fixes the issue....

A niche issue but when running task 320 with a scissor correction with a non-ground-state set of occupations, then: In genvchi0, if the conduction band has occupation, then transitions within the conduction bands will erroneously get a scissor correction. This can lead to a spurious peak at the scissor correction energy. In task 121, the dielectric routine doesn't have this problem as it checks whether the transition goes from valence to conduction. Copying this code to genvchi0 fixes the issue....