(wrong) phonons with r2SCAN

[Tyler Sterling]

Hi,

I am trying to model La2NiO4 using r2SCAN. I am using r2SCAN because I also want AFM order, which GGA typical doesn't get right. For now, I am calculating the "high-temperature" paramagnetic I4/mmm structure as a test. I am calculating phonon dispersions. This isn't the actual ground state.. so I do expect certain unstable modes. With PBE, the dispersions look reasonable with the "right" instabilities. With r2SCAN, there are additional unstable modes that shouldn't be unstable. I tried increasing CUTOFF (from 2000 Ry up to 2800 Ry!) and tried with "SCAN" pseudo-potentials as well as with PBE pseudo-potentials and found even more unstable modes! I am no expert in SCAN calculations or with CP2K, so I am worried I am doing something wrong in the input file. The other possibility is that my supercell (3x3x1 expansion of the conventional I4/mmm cell) for phonons (using phonopy) isn't large enough: I read somewhere that SCAN functionals are much more "long-ranged", so maybe I need even bigger to converge. 

I am attaching some minimal r2SCAN and PBE input files. These were both relaxed to within max-force=0.0001 Ry/Bohr before phonon calculations (all other relaxation params left at default values). 

I got the La basis set from the BASIS_MOLOPT_UCL file and the SCAN potentials from the POTENTIAL_UZH file. All other basis sets and potentials are from BASIS_MOLOPT and POTENTIAL respectively.

Does anyone have any idea if I am doing something wrong here? Any help is appreciated! Thanks :)

Ty