Closed eminamitani closed 3 years ago
I have tested anisotropic cell for SiO2 but it seems OK. The origin of the problem is not sure...
Strange large diffusivity is not from cell. That is caused by 1/omega contribution for artificial small frequency mode. As tentative fix, I set the threshold in get_Sij as
inv_omega=np.zeros(nmodes)
for i in range(nmodes):
#tentative
if omega[i] >0.1:
inv_omega[i]=1.0/np.sqrt(omega[i])
else:
inv_omega[i]=0.0
This is not so good way. The better solution is acoustic sum rule to avoid negative frequency
In some test, the result of thermal conductivity is overestimated. This may due to some bug in the nearest distance. Thus, try calculation in non-cubic cell and check the value of the nearest distance.