I think the first criteria is that dx should not be so small that the resulting dE become comparable or smaller than energy accuracy of your system. Or other way around, the energy accuracy of your calculation should be considerably larger than dE, it means that if you reduce dx, then accuracy (kpoints, rkmax, convergency) should be increased consistently.<br>
<br>Javad Hashemifar<br><br><div class="gmail_quote">2009/7/14 Atta-fynn, Raymond <span dir="ltr"><<a href="mailto:attafynn@uta.edu">attafynn@uta.edu</a>></span><br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div>
<div dir="ltr"><font color="#000000" size="3" face="Tahoma">Dear WIEN2k users and authors,</font></div>
<div dir="ltr"><font face="tahoma"></font> </div>
<div dir="ltr"><font face="tahoma">As it stands, the Hellmann-Feynman forces cannot be computed when spin-orbit coupling (SOC) is present. I want to compute the phonon dispersion relations for an fcc cell using PHONON but I want to use SOC forces so I intend
to compute the forces on the atoms using the SOC total energies. The force on a single atom along a coordinate direction, say x, will be calculated as </font></div>
<div dir="ltr"><font face="tahoma"></font> </div>
<div dir="ltr"><font face="tahoma"><font face="tahoma">Fx_i</font> = del_E/(2*dx), where del_E = Etot (x_i + dx) - Etot (x_i - dx)</font></div>
<div dir="ltr"><font face="tahoma"></font> </div>
<div dir="ltr"><font face="tahoma">Following the above procedure for an fcc cell with 4 inequivalent atoms, 8 total energy calculations will be carried out for each structure (assuming I want forces along the x-direction only).
</font></div>
<div dir="ltr"><font face="tahoma"></font> </div>
<div dir="ltr"><font face="tahoma">My concern is how to choose a "good value" of dx that will yield forces accurate to within 0.1 mRy/Bohr. My guess is dx = 0.01 Bohr but I would like to seek some input from other users before I proceed with the calculations. Also,
should the total energy be computed to a higher degree of accuracy, say 8 or 9 decimal places? </font></div>
<div dir="ltr"> </div>
<div dir="ltr"><font face="tahoma">Thank you.</font> </div>
</div>
<br>_______________________________________________<br>
Wien mailing list<br>
<a href="mailto:Wien@zeus.theochem.tuwien.ac.at">Wien@zeus.theochem.tuwien.ac.at</a><br>
<a href="http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien" target="_blank">http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien</a><br>
<br></blockquote></div><br><br clear="all"><br>-- <br>======================================<br>Seyed Javad Hashemifar<br>Physics Department, Isfahan University of Technology<br>84156-83111 Isfahan, Iran<br>Tel: +98 311 391 2375 Fax:+98 311 3912376<br>
Email: <a href="mailto:hashemifar@cc.iut.ac.ir">hashemifar@cc.iut.ac.ir</a><br>Homepage: <a href="http://hashemifar.iut.ac.ir">http://hashemifar.iut.ac.ir</a><br>---------------------------------------------------------------------------<br>