[lammps-users] Regarding the Green-Kubo method in lammps

HI steve and all,

Thanks for bringing the new compute heat/flux to work. However, I found some issues as follows:

  1. Typo:
    reset_timestep 0
    compute myKE all pe/atom
    compute myPE all ke/atom

should be:
reset_timestep 0
compute myKE all ke/atom
compute myPE all pe/atom

  1. in the Green-Kubo equation on Lammps website document:
    k = v/(k_B*T^2)*integral of heat auto-correlation

the volume should be divided by the other terms not be multiplied by (see lots of Green-Kubo method publications).

  1. IN the LJ example,

V = 10213.257 Angs^3

KB = 1.38e-23 (J/K) = 1.38*6.022111e-3/4.182 = 1.98721e-3 (Kcal/mol K) (instead of 1.98816 KCal/(mol K))

J_intergral = 9e-11

lamda = J_integral* output frequency * timestep/(VolK_BT^2) = 9e-11104/(10213.2571.98721e-370*70)= 3.61992E-14

For above, I used consistence unit: REAL units. So the lamda is in unit of Kcal/(mol fs A K) (if lammps give does not do anything wrong)

the convert factor = 4.182e3 / (1e-15 * 1e-10 * 6.022e23)) = 69478.6

Kapar = lamdaconver factor = 3.61992E-1469478.6 = 2.5E-9 (W/mK)

Can anybody verify my solution?