Matt. I've attached the output files for my runs. I tried OT in 3 ways, first restarting from the diag wfn, second using the atomic guess (It shows a restart, but restarted from unconverged OT). The atomic guess was unable to converge, but the unconverged result shows the same negative band gap. The input for the OT run is also attached.<div><br></div><div>Thoughts?<br><div><br></div><div>-Nick</div></div><br><div class="gmail_quote"><div dir="auto" class="gmail_attr">On Thursday, June 3, 2021 at 7:24:31 AM UTC-7 mat...@gmail.com wrote:<br/></div><blockquote class="gmail_quote" style="margin: 0 0 0 0.8ex; border-left: 1px solid rgb(204, 204, 204); padding-left: 1ex;">Hard to say without seeing more. I guess it has converged to the wrong electronic state somehow.<div>Did you restart from the Diag wfn file? You could try starting from an atomic guess instead if so and see what that produces.</div><div>That the LUMOs converge in one iteration seems suspicious somehow.</div><div>Matt<br><br></div><div class="gmail_quote"><div dir="auto" class="gmail_attr">On Wednesday, June 2, 2021 at 8:25:40 PM UTC+1 <a href data-email-masked rel="nofollow">n...@berkeley.edu</a> wrote:<br></div><blockquote class="gmail_quote" style="margin:0 0 0 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">Hello everyone. I have a semiconductor, BeSiP2, which has a band gap slightly above 1eV in the GGA approximation (<a href="https://materialsproject.org/materials/mp-1009085/" rel="nofollow" target="_blank" data-saferedirecturl="https://www.google.com/url?hl=en&q=https://materialsproject.org/materials/mp-1009085/&source=gmail&ust=1622825807106000&usg=AFQjCNHXEOa_wuGN7zpJpbgNcCcy76pQAg">https://materialsproject.org/materials/mp-1009085/</a>).<div><br></div><div>I calculate a vacancy, V_{P}^{+3}, first using diagonalization. Which produces a band gap around 1eV, indicating that the charge does not delocalize, and so OT should still be appropriate. When I switch to OT, however, there is a negative homo/lumo gap and I can see it in the eigenvalues:</div><div><br></div><div><p>0.17131297 0.17154145 0.18362819 0.18519005</p>
<p> Fermi Energy [eV] : 5.039278</p>
<p> Lowest Eigenvalues of the unoccupied subspace spin 1</p>
<p> -----------------------------------------------------</p>
<p> Reached convergence in 1 iterations</p>
<p> 0.18499549 0.22295666 0.23850766 0.25228060</p><p><br></p><p>This is bizarre to me. OT is usually very reliable for me. So this is a strange inconsistency, and maybe its due to how I'm running the calculation. For OT I use the full_all preconditioner, and the rest defaults. For diagonalization I the section at the bottom. Anyone have any guesses as to why there is this discrepancy? </p><p>-Nick</p><p>&SCF</p><p> MAX_SCF 200</p><p> added_mos 100</p><p> eps_scf 1e-6</p><p> max_diis 15</p><p> &diagonalization</p><p> &end</p><p> &mixing</p><p> &end</p><p> &smear</p><p> elec_temp 300</p><p> method fermi_dirac </p><p> &end</p><p>&end</p></div></blockquote></div></blockquote></div>