Calculating diffusion coefficient with MSD (gmx msd)

GROMACS version: 2016

Hi all,

I’m trying to calculate the diffusion coefficient of some water molecules in the x direction of my system. I used gmx msd and did the following:

gmx_mpi msd -f nvt.trr -s nvt.tpr -n index.ndx -mol -b 1000 -e 1900 -beginfit 100 -endfit 300 -type x

which GROMACS automatically calculates the diffusion coefficient, D[ Water] 1.6535 (+/- 0.1480) 1e-5 cm^2/s. However, when I manually calculate the diffusion coefficient (over the linear region between 100 and 300 ps which I told GROMACS to use as well) by calculating the slope of my msd plot, I get 0.0031806 nm^2/ps (3.1806e-5 cm^2/s). Does anyone know what might be happening? It looks like I’m around a factor of 1/2 out from what GROMACS gives but I don’t know why.