Difference in diffusion coefficients using msd and velacc functions

User discussions
1

GROMACS version:2024.1
GROMACS modification: No
For the same generated phase, the diffusion coefficients produced using gmx msd and velocity autocorrelation are significantly different, how should one determine which is accurate?
For a 1M MgCl2 solution, the system uses 500 water molecules, 9 Mg and 18Cl.
Calculate the diffusion coefficient for the Mg ion:
Using gmx msd to select Mg, the diffusion coefficient calculated directly is: @ s0 legend “D[ Mg] = 0.2957 (+/- 0.2086) (1e-5 cm^2/s)”
And using gmx velacc and gmx analyze -integrate using the velocity autocorrelation function integrated and divided by 3 comes out to 0.008231067 cm^2/s

Analysed quantitatively they are both very different, what is the reason for this and which one is correct?

2

What kind of simulation system are you addressing? time length of the simulation? pbc is fixed? both method uses different approaches and it varies on several factors.

3

thank you for answer my question, and in my case, I use tcoupl = V-rescale, pcoupl = C-rescale , the nsteps is 10000000, dt =0.02 , use Annealing Simulation start 500K to 298.15K and pbc =xyz in eql.mdp. and i have one more question is what is units of gmx current outputs value, sigma? is S/m?