<html><head></head><body dir="auto" style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;"><div class="ApplePlainTextBody"><br>Dear all,<br><br>I currently run slab calculations including spin-orbit coupling for around 30-60atoms large systems. 3-4 different species involved; all norm-conserving pseudo-potentials. Self-consitency runs fast & smoothly into the equilibrium. <br>The subsequent non-scf (bands) run crashes however with the infamous error from cdiagh „problems computing cholesky“. <br>That puzzles me. I don’t have ghost states in my PP’s (as far as I checked); furthermore they are NC as noted - I thought that the cholesky decomposition problem shouldn't appear here. <br>I changed to the more stable conjugate gradient method. It surprisingly still fails. <br>I noted that decreasing the number of bands to something only slightly larger than the occupied bands prevents the error to pop up. Unfortunately I’ll need quite a lot of unoccupied states for my post-process, so nbnd will be large (at least 400bands for the smallest slab.)<br><br>Is there a known relation of the number of bands and the error occurring during the Davidson diagonalization. Anyone had the same problem?<br><br>thanks a lot in advance,<br>Nicki<br><br>CAMd, DTU Physics</div></body></html>