Imaginary phonon modes appearing at high temperatures in SCPH calculations

[kamalwahhid-cyber]

Hi,
I am performing SCPH calculations and the self-consistent loop converges over the whole temperature range (0–1000 K). However, I observe that imaginary phonon modes start to appear above ~700 K, while the system is stable at lower temperatures.
How should this behavior be interpreted?
Does it indicate a real high-temperature dynamical instability (e.g., a phase transition), or could it be related to inaccuracies in the higher-order interatomic force constants (especially 4th order) or SCPH convergence issues at high temperature?
Any insights or similar experiences would be very helpful.

[ttadano]

If the imaginary mode appears at interpolated q points (off from the uniform grid of KMESH_INTERPOLATE), I recommend that you increase KMESH_INTERPOLATE and use harmonic IFCs computed using the same size of KMESH_INTERPOLATE.

As you noted, inaccurate IFCs can also make this happen, so I recommend that you check if the behavior remains/changes after changing the number of training data used for extracting anharmonic IFCs.