I'd like to replicate the results from this paper: https://doi.org/10.1103/PhysRevB.80.165203But I've had some issues as of now. The paper describes their set up as: 'A fully relaxed structure has rhombohedral unit-cell parameters a_R = 10.27 Å,theta_R = 24.49° and atomic parameters u = 0.400, v = 0.791". I use a lattice I downloaded from pymatgen with similar parameters, make a (6, 6, 1) super-cell (also 540 atoms) and then let it relax NPT, NVT and NVE before outputting HACF for post-processing. I use LAMMPS for all MD runs. One thing I notice is that my structure goes from 26.72 nm^3 to ~ 17.8 nm^3. I don't know if such a collapse is reasonable (the QL layers seem fine), but regardless the compute HACF looks like pure noise, as opposed to a decaying signal as shown in the paper. I computed the HACF for a simple LJ model and compared the thermal conductivity value to a reference and it was within error of one run so I have no worries about my HACF method. Would someone be able to look at my code for structure generation/relaxation and give me some pointers? Thank you!
↧