Nelson David Rufus

Hello, I was doing a system size study on Cu supercells (Gamma-pt). I started off with a 4 atom FCC Cu unit cell and created supercells by translating this unit cell. The groundstate energies per atom (Ha) are tabulated below: System | ELK | In-house Finite Element Code 1x1x1 (4 atoms) | -1,637.8639 | -1637.8634 2x2x2 (32 atoms) | -1,622.7617 | -1637.9275 3x3x3 (108 atoms) | (in progress) | -1637.9301 Clearly, there is something wrong with the 2x2x2 ELK value. I used an rgkmax=9.0, which is the same...

Hi Kay, I reproduced the Uranium case with my Elk code. Moreover, since the basis of my Finite Element code is local, I did an isolated Ce atom in a large (80 au) box and found there were some disagreements with dftAtom for the same occupancies and xc functional. I later found an error in my Finite Element calculation which I have now fixed. After that, dftAtom agreed with my isolated atom calculation. As of now, the 4-atom Ce Gamma-pt calculation with Finite Elements and Elk show good (~0.1-0.2...

Hi, Thanks for getting back. Just to clarify one thing, the calculation that I was referring to earlier was a gamma-pt calculation. I tried the vhighq flag and I noticed that it automatically picked a refined kpoint grid. But even in this case, when I compare with my (kpoint converged) Finite element calculation, there is an error of ~6 mHa. The NIST calculations use a different LDA (Vosko, Wilk, Nusair) which I don't have with my code at the moment. So I can't test it right away. When I was going...

Hi Kay, In the Finite Element formulation, we treat core and valence electrons on the same footing i.e. the first N eigen values of the dircrete generalized eigenvalue problem will cover both the core and the valence bands. So there is no need to calculate the core eigen values separately using a radial Schrodinger equation. Also, my calculation is non-relativistic. Hence I assign solscf = 1000 when I try the equivalent Elk calculation. Regards, Nelson

Hello, I performed a non-spin polarized groundstate calculation (sample input file attached) for FCC Cerium (4 atom cubic simulation cell). I checked convergence with respect to rgkmax and compared it against an inhouse Finite element calculation. I observed an error of around ~ 5mHa in the groundstate energy per atom. I looked up the Elk documentation and also checked against other parameters like lmaxi , lmaxo, lmaxapw, solscf (non-relativistic calculation), nempty, rminsp, nrmt but the Elk solution...

Thank you very much!!

Hello, I assume that the default setting involves relativistic-KS Hamiltonian. Is there a way to switch it OFF and only perform a non-relativistic calculations. (For example, "exciting" dft has a couple of flags to switch off the coreRelativity and valenceRelativity). Is there an equivalent in Elk. I tried searching in the manual but couldnt find any such flag. Thanks, Nelson

Thanks for the information. I'll try it out.