Error in routine cdiaghg (146): S matrix not positive definite in d3q.x calculation

[minnie178]

Hi,
The d3q.x calculation for SnO is giving an error
"Error in routine cdiaghg (146):
S matrix not positive definite" in the output file.
The phonon calculations completed without any issue. Could it be any problem with the pseudopotentials? ONCV pseudopotentials are used.
The scf, phonon,
and d3q.x input files and corresponding outputs are uploaded here for reference.
d3q_SnO.zip
Any help on this will be greatly appreciated. Thanks in advance!

[paulatz]

The error is raised during the nscf calculation that precedes the actual d3q calculation, in your pw.x input, you are using

  ecutrho =   480
  ecutwfc =   60

I.e. ecutrho=8ecutwfc, this is never necessary when using norm-conserving pseudopotentials, where the default value ecutrho = 4ecutwfc should always be used. The d3q code, as it has never been tested under these conditions, there is a chance some bug is present, I do not see anything else problematic in your input so I would test using the default ecutrho first. Maybe you could also tighten the phonon convergence, tr2_ph a little bit to 1.d-16, this won't solve any bug but may help with the results.