Fig. 16. Workflow to compute the bulk limit of Allen-Feldman and rWTE conductivities.

The amorphous models employed in this work were generated (step 1) using the melt-quench method (see Sec. Calculations performed using the GAP potential for the 1536(G) model, ref. ⁶⁰ for 192(G) and 5184(G), ref. ⁶³ for 144(D)) or the bond-switching technique (see ref. ⁶² for 108(D), and ref. ⁶⁵ for 192(D)). After relaxing the atomic position and reference cell to an energy minimum, we computed the harmonic (step 2a) and anharmonic third-order (step 2b) force constants. Then, harmonic force constants were used to compute the vibrational frequencies and velocity operator over a computationally converged q mesh (step 3a), thus the convergence plateau for the Allen-Feldman conductivity was determined (see Fig. 2). The broadening η determining the beginning of the convergence plateau was used to compute the Allen-Feldman conductivity (step 4). Finally, harmonic and anharmonic force constants were used to compute the linewidths (step 3b), and these were used in combination with the aforementioned broadening η to evaluate the rWTE conductivity (step 5, see Eq. (1).