stress tensor from MD simulation using CHGNET

[gaoxuew]

Dear developers,
I am interested in calculating viscosity using off-diagonal elements of stress tensor from MD simulations using CHGNET. Could you please comment how good are the off-diagonal elements of stress tensors from CHGNET? Is the ideal gas correction included in the output?
Thank you.

[BowenD-UCB]

I have not tested the off-diagonal parts of stress tensor explicitly, but here is my comment, I hope it's helpful.

chgnet.predict_structure does not include the ideal gas correction. But when running the MD in ASE, the ideal gass correction is added here, which is calculated from the momentum carried in ASE atoms.

ASE MD also assume the stress tensor is symmetric by default.

If you have any benchmark result on the off-diagonal term, please share with us as it will be very helpful.