My excuses, I strongly disagree with Roberto's opinion: > > Hi akshu, > >> dear siesta users >> i need to calculate the bulk modulus in order to test the >> pseudopotential >> for technetium. Tc is found in hexagonal structure, so along with >> varying >> the lattice constant, do i need to change 'c' as well? > > The closest thing to bulk modulus in hcp is obtained by varying > all the cell dimensions simultaneously and homogeneously and > monitoring the pressure while doing that. Then P = B \delta V / V.
The definition of bulk modulus (B) known to me is B = -V dP/dV. As the structure is not cubic, varying cell dimensions simultaneously will introduce stress (the relaxed cell won't like to have the same c/a for different volumes). I think the right thing to do is to vary TARGET PRESSURE, let cell parameters free and monitor the VOLUME, then use the formula above. Or - probably better - you extract elastic constants, checking back their definition in hcp, as indepent parameters. Because they may behave differently; the bulk modulus is just some averaged combination of them, and won't tell you much, for a serious test. >> is it enough to use md.variable cell? how will i know that cell geometry >> remains hexagonal? > > Of course you shouldn't use variable cell: for each \delta V above the > cell is kept fixed, otherwise the pressure will relax to zero. Sorry again, I thing, it is ESSENTIAL to use variable cell, for the reason given above. >> should i fix the atomic positions? > > No. The hcp 2-atoms cell possesses inner degrees of freedom, How's that??? I always thought there is something like (0 0 0) and (1/3 2/3 1/2), no internal coordinates. These relative coordinates should not change (they will in fact, slightly, due to lack of symmetry constraint in Siesta). > most likely the symmetry will be broken. > >> Is it all right if i test the pseudo for fcc structure instead and >> compare >> the results with available theoretical results. And what you want to compare with otherwise, the experiment? Is there some, for technetium? Just wondering. > Who knows. Any test is just a test, it should be helpful anyway. Hmm. I think, having a good all-electron code at hand to compare, you can, to begin with, construct any distortion you want, do the same with Siesta, compare one curve against the other and already get an idea. For this you don't even need to know what the bulk modulus is. Best regards Andrei
