Dear Andreas, I'd suggest you first to check whether the band structure, and not just the band gap, comes out right. Try to have a look at least at PDOS. The point is, NiMnSb has some magnetic structure, which would hardly emerge by itself by default. You define spin polarization = true in your input file but do not elaborate further. In this case (see the Siesta manual), all atoms (including Sb) are initialized with maximal magnetic moment, all parallel. The chances are, such initial configuration will converge, but to something completely unreasonable. You can get an idea of the resulting magnetic structure by switching on mulliken occupations, Probably it converges to different magnetic structures at different lattice constant, hence the "ugly jumps". Another possible reason for discontinuities in the Etot(a) curve may be a too low k-grid cutoff, so that the actual (low) k-mesh number of divisions changes at some values of a.
Best regards Andrei Postnikov > > <div>I'm a new user to siesta and want to determine the equilibrium > lattice constant for the half-metallic heusler alloy NiMnSb. To do this I > downloaded the GGA pseudopotentials for Ni and Mn and created my own for > Sb with the atom program. My problem is that either I have no band gap at > the fermi Energy in the minority spin domain , or the plot total Energy > vs. lattice constant has ugly jumps. I could not yet achieve a nice > parabolic curve for the lattice constant with a band gap at the minimum > Value. I'll attach my input file. If any other files are important to > you just let me know. I'd be glad if anyone can help me out.</div> > > <div> </div> > > <div>greetings, Andreas Prinz-Zwick</div></div></body></html>
