Hi, I am trying to find the self-diffusion coefficient of 3000 water molecules using different forcefields. Of all the tested forcefields, universal forcefield gave me the value (2.4 x 10^-9m^2/s) closest to the experimental value. However, the value keeps drastically changing. When I setup a simulation for 1ns production run, it randomly terminated at 50ps and analysis of the frames thus far gave me the desired D.C result. No errors were registered in the project log. I then set up another simulation with same settings and from the same input structure for 50ps, and the result at the end of simulation was 36.540 x10^-9m^2/s, far higher than the desired value. I then decided to set up the production run for 1ns as I earlier intended and analysed the frames midway at 50ps while the job continued. At that time I got the desired value again (2.4 x 10^-9m^2/s) but once the job completed after 1ns, the D.C was (37.528 x 10^-9m^2/s) when I analysed the frames upto 50ps and up to 1ns. For all the simulations the density drops from 1g/cm^3 to 0.4g/cm^3 during the production run as shown in the photo. I need help understanding why the D.C is changing and why the density is dropping that much.