<html><head></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; ">Dear Murat,<div><br></div><div><br><div><div>On Jan 28, 2011, at 4:53 PM, Murat KILIC wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite"><div><div style="font-family: Helvetica; font-size: medium; "><br></div><div style="font-family: Helvetica; font-size: medium; ">Dear cp2k developers and users,</div><div style="font-family: Helvetica; font-size: medium; "> <br></div><div style="font-family: Helvetica; font-size: medium; ">We have been performing now for some time mixed hamiltonian MD simulations of an organic molecule in water at the DFT/BLYP level of theory, and in some simulations we noticed some very awkward behavior: that is, sometimes one or more atoms suddenly completely stop moving.</div> <div style="font-family: Helvetica; font-size: medium; "><br></div><div style="font-family: Helvetica; font-size: medium; ">We have done various tests to understand this problem. We have now even seen this behavior in a rather standard MD run of 32 water molecules at T=350 K using the BLYP XC functional and the TZV2P basisset. Although we start from a restart file was generated in a special manner (using a strange mixed hamiltonian), this example here is a rather typical simulation which only uses coordinates and velocities from a restart file.</div></div></blockquote><div><br></div><div>When something like this happens it just means that there is some numerical problem in your calculations. </div><div>Most of them can normally be related to the cutoff, generally too small to have a good representation either of the pseudo or the basis set.</div><div>You have 350, which is a pretty good number both for BLYP and the kind of pseudo. Maybe for the basis set you would need something higher, but still, it should be not that bad.</div><div><br></div><div>What I think is the major source of this problem in your case is the convergence of the SCF. 10^-5 (default) with OT/DIIS is a too loose convergence criteria to have pretty good converged forces. If you think at the issues connected with the cutoff (mentioned above) you realize that your forces, in some cases, for some geometries, maybe completely covered by numerical noise.</div><div><br></div><div>Try to decrease the EPS_SCF to 10^-6 or tighter (even 10^-7) and let us know if you still see this behavior.</div><div><br></div><div>Regards,</div><div>Teo</div><br><blockquote type="cite"><div> <div style="font-family: Helvetica; font-size: medium; "><br></div><div style="font-family: Helvetica; font-size: medium; ">We have put an example of 1000 MD steps online at: <a href="http://molsim.chem.uva.nl/users/ensing/cp2k_problem/water32/">http://molsim.chem.uva.nl/users/ensing/cp2k_problem/water32/</a></div> <div style="font-family: Helvetica; font-size: medium; ">In the trajectory file the oxygen atoms with index 4 and 61 (start counting from 1) are not moving (actually the positions fluctuate in the 3rd decimal in all our examples).</div> <div style="font-family: Helvetica; font-size: medium; "><br></div><div style="font-family: Helvetica; font-size: medium; ">Has anyone seen this problem before? Is this a bug in the code or are we doing something wrong?</div> <div style="font-family: Helvetica; font-size: medium; ">We know that we can "release" the atoms from their apparent fixed positions by changing input parameters. For example, increasing the planewave cutoff seems to reduce the number of atoms that typically get stuck in our simulations. Changing computer platform or cp2k version does not matter.</div> <div style="font-family: Helvetica; font-size: medium; "><br></div><div style="font-family: Helvetica; font-size: medium; ">Any ideas are welcome.</div><div style="font-family: Helvetica; font-size: medium; ">Many thanks in advance,</div> </div><div style="font-family: Helvetica; font-size: medium; "><br></div><div style="font-family: Helvetica; font-size: medium; ">Murat</div><div><br></div><div><br></div><div><br></div><br clear="all"><br><br>Murat KILIC<br> Computational Chemistry Group<br>Van 't Hoff Institute for Molecular Sciences<br> University of Amsterdam<br> Science Park 904<br> 1098 XH Amsterdam<br><a href="http://molsim.chem.uva.nl/users/kilic/index.html" target="_blank">http://molsim.chem.uva.nl/users/kilic/index.html</a><br><br><br><div><br class="webkit-block-placeholder"></div> -- <br> You received this message because you are subscribed to the Google Groups "cp2k" group.<br> To post to this group, send email to <a href="mailto:cp...@googlegroups.com">cp...@googlegroups.com</a>.<br> To unsubscribe from this group, send email to <a href="mailto:cp2k+uns...@googlegroups.com">cp2k+uns...@googlegroups.com</a>.<br> For more options, visit this group at <a href="http://groups.google.com/group/cp2k?hl=en">http://groups.google.com/group/cp2k?hl=en</a>.<br> </blockquote></div><br></div></body></html>