Hello, I am attempting to calculate the q=0 phonon modes for a system comprising a CO molecule above a Au(111) surface. The input files for both the pw.x (slab1.pwi) and the ph.x (slab1.phi) calculations as well as their corresponding output files (slab1.pwo and slab1.pho). These calculations were run with QE 6.5 compiled with the Intel compiler stack (ifort/MKL/IMPI), and the jobs were executed on 5 compute nodes of 36 CPU cores each. Both pw.x and ph.x were run with the “-nk 5” argument to enable k-point parallelization. Note that the pw.x input file was generated with ASE, and some of the settings like ecutrho and ecutwfc are much rounder numbers when represented in eV rather than Rydberg.
The problem I am encountering is that the ph.x calculation does not converge even for the first representation/mode. In fact, the calculation *diverges*, with |ddv_scf|^2 = 2.464E+18 at the 100th iteration. I am wondering whether I have done something wrong, or if there are some keywords that should be tweaked to improve convergence. I have seen advice on this mailing list to lower the value of alpha_mix(1) from its default value of 0.7, but if I understand the manual correctly this would result in using more new density, which I imagine would make convergence even worse in my case. I haven’t been able to find any advice online for improving convergence of metallic systems. Regards, -- Eric Hermes Postdoctoral Researcher Sandia National Laboratories
<<attachment: gold.zip>>
_______________________________________________ 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
