<html><head><meta http-equiv="Content-Type" content="text/html charset=utf-8"></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space;">I find that tightening the convergence criterion on various evaluations generally solves the low frequency hessian convergence problems (my experience is that anything below about 200 cm-1 should not be considered reliable with the standard parameters).<div>Thus, I generally insert the following line at the top of my input file</div><div><div style="margin: 0px; font-size: 10px; font-family: Monaco;">gthresh,energy=1.0d-10, orbital=1.0d-10, oneint=1.0d-16, twoint=1.0d-16, optgrad=1.0d-6, compress=1.0d-13</div><div>Best Regards,</div><div><span class="Apple-tab-span" style="white-space:pre"> </span>Stephen<br><div><div>On Feb 6, 2017, at 9:12 AM, Leonid Shirkov <<a href="mailto:leonid.shirkov@gmail.com">leonid.shirkov@gmail.com</a>> wrote:</div><br class="Apple-interchange-newline"><blockquote type="cite">Dear Colleagues,<br><br>in some specific cases, the current accuracy of the numerical hessians<br>is not enough, e.g. for very low frequency torsional vibrations (~50cm-1).<br>The CCSD(T) freq analysis gives imaginary values instead of the real ones<br>for the lowest modes. The solution is to find manually the hessian in<br>the internal coordinates<br>and then find the eigenvalues of the GF matrix, but that is a lot of work.<br><br>If MP2 is used for such cases, then there are no imaginary frequencies.<br>Do I understand correctly, that for MP2 freq analysis the hessians are found<br>by differentiating the analytical MP2 gradients?<br><br>Using the analytical gradients for highly accurate methods like<br>CCSD(T) would probably resolve the problem,<br>but they are not currently available in Molpro.<br><br>Best regards,<br>Leonid<br><br>On Mon, Feb 6, 2017 at 10:18 AM, Werner Győrffy<br><<a href="mailto:gyorffy@theochem.uni-stuttgart.de">gyorffy@theochem.uni-stuttgart.de</a>> wrote:<br><blockquote type="cite">Dear Aleksandr,<br><br>Numerical Hessians in Molpro are computed by using central finite<br>differences with a 2-point formula as a default. That is a "general<br>formula". That gives accurate results in most of the cases. There is a<br>trade-off between accuracy and efficiency: More accurate finite field<br>calculations would increase the number of single point calculations<br>significantly. If one needs more accurate Hessians, it can be done only by<br>computing that manually by using procedures.<br><br>Regards,<br><br>Werner.<br><br><br>On 02/04/2017 02:04 AM, Aleksandr Lykhin wrote:<br><blockquote type="cite"><br>Does anybody know how Molpro calculates mixed derivatives using central<br>differences? It seems like it generates only two mixed displacements<br>instead of four, so the general formula cannot be applied directly.<br><br>--<br>Kind regards, Aleksandr O. Lykhin.<br><br><br>_______________________________________________<br>Molpro-user mailing list<br><a href="mailto:Molpro-user@molpro.net">Molpro-user@molpro.net</a><br>http://www.molpro.net/mailman/listinfo/molpro-user<br><br></blockquote>_______________________________________________<br>Molpro-user mailing list<br><a href="mailto:Molpro-user@molpro.net">Molpro-user@molpro.net</a><br>http://www.molpro.net/mailman/listinfo/molpro-user<br></blockquote>_______________________________________________<br>Molpro-user mailing list<br><a href="mailto:Molpro-user@molpro.net">Molpro-user@molpro.net</a><br>http://www.molpro.net/mailman/listinfo/molpro-user</blockquote></div><br></div></div></body></html>