Hi QE Experts, I have a system whose ground state is a doublet ( 1 spin up electron, in this case ). I am using ultrasoft PPs. The system is a 3x3x3 super cell of diamond with a single NV center.
My strategy for converging Ecutwfc and Ecutrho follows the prescription laid out by Stephano in previous exchanges on the forum. It involves two steps: 1) Using the default Ecutrho = 4*Ecutwfc, I converge the total force of my cell w.r.t Ecutwfc. 2) Then, using the converged value of Ecutho, I scale back Ecutwfc to see how low I can get it without coming out of my tolerance window, which is 1mRy in the total force computed using the converged parameters. When I make my unit cell charged (therefore a triplet ground state [2 spin up electrons] ) - I get a great convergence. However, when doing step1 using my charge neutral (doublet) ground state, I obtain the following for step 1 in the convergence procedure: 20.0000 0.238517 30.0000 0.273459 40.0000 0.272399 50.0000 0.270205 60.0000 0.269809 70.0000 0.269807 80.0000 0.274589 90.0000 0.274698 100.0000 0.270163 110.0000 0.270168 120.0000 0.270176 130.0000 0.270245 140.0000 0.274755 150.0000 0.270246 The left column is Ecut (Ry), the right column is Total Force (Ry/Bohr). One can notice "spikes", for example, at Ecut = 140, and E cut = 80-90 Ry. Can such a thing happen for a particular difficult occupation scheme? I suspect that I am oscillating between two solutions, and that using Ecut 50 (therefore Ecutrho = 200 Ry) works here. Then, using Ecutrho = 200, I can most likely get the Ecutwfc back to ~ 35 or so. Thank you for your time, Best, Brendan A. Smith 4th year graduate student studying computational chemistry at the State University of New York at Buffalo
_______________________________________________ Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso) users mailing list [email protected] https://lists.quantum-espresso.org/mailman/listinfo/users
