<p dir="ltr">A slightly tangential comment: unless your CIF came from a neutron diffraction analysis the H positions can be very wrong. Many refinement codes such as Shelx are very good, but just put H into "reasonable" positions as they are unstable in the refinements with XRD data.</p>
<p dir="ltr">In principle the CIF file should have information about how the H positions were determined.</p>
<p dir="ltr">___________________________<br>
Professor Laurence Marks<br>
Department of Materials Science and Engineering<br>
Northwestern University<br>
<a href="http://www.numis.northwestern.edu">www.numis.northwestern.edu</a><br>
<a href="http://MURI4D.numis.northwestern.edu">MURI4D.numis.northwestern.edu</a><br>
Co-Editor, Acta Cryst A<br>
"Research is to see what everybody else has seen, and to think what nobody else has thought"<br>
Albert Szent-Gyorgi<br>
On Mar 20, 2015 6:56 AM, "Peter Blaha" <<a href="mailto:pblaha@theochem.tuwien.ac.at">pblaha@theochem.tuwien.ac.at</a>> wrote:<br>
><br>
> Of course, for "better" positions it will be easier, but you would still<br>
> run into the same problems with the default inputs.<br>
><br>
> The problem with ghostbands is either connected with<br>
> i) a WRONG case.struct file (not in your case) or<br>
> ii) ALWAYS connected with the VERY SMALL spheres (for atoms like C,N,O<br>
> or in your case P and O) AND the s- (l=0) states of these atoms.<br>
><br>
> Typically the 3s (2s) state of P or O (or C,N) is lower in energy than<br>
> the actual valence states and for a BIG sphere it is therefore necessary<br>
> (more accurate) to describe these states with an additional LO and 2<br>
> lines in case.in1 like:<br>
><br>
> 0 -1.55 0.002 CONT 1<br>
> 0 0.30 0.000 CONT 1<br>
><br>
> For SMALL spheres, however, this leads to ghostbands and the case.scf2<br>
> file indicates for you for which atom and l it happens (atom 3 for l=0).<br>
><br>
> Thus you have to edit case.in1 and either set the "second" energy very<br>
> high (O):<br>
><br>
> 0 -1.55 0.002 CONT 1<br>
> 0 5.30 0.000 CONT 1<br>
><br>
> or even remove the second line (and change the number of "exceptions" in<br>
> the line above from 3 to 2 (P-atom):<br>
><br>
> 0.30 2 0 (GLOBAL E-PARAMETER WITH n OTHER CHOICES,<br>
> 0 -1.79 0.002 CONT 1<br>
> 1 0.30 0.000 CONT 1<br>
><br>
> Then do a "crude":<br>
><br>
> run (-p) -fc 10<br>
><br>
> and you see enormous forces on many atoms (in particular H, because<br>
> their distance is VERY wrong, leading to these extremely small H and O<br>
> radii).<br>
><br>
> run -p -min -fc 1<br>
><br>
> optimizes the positions seamlessly. I've stopped this at a very early<br>
> stage, but the struct file is attached.<br>
><br>
> With this struct file I would rerun setrmt again, because the much more<br>
> realistic O-H distances allow now for larger H and O spheres making the<br>
> calculations MUCH faster. Then continue with the position optimization.</p>