I am trying to calculate the viscosity of fused silica at different temperatures (300 K, 400K, 500K, … 4000K). I am using the viscosity calculation example from lammps.org, but however don’t I change the parameters (correlation length, sample interval, dump interval), the viscosity is always low (near the range of 0.01 Pa.s), while in reality at low temperatures the viscosity of silica is about 10^18 Pa.s). So my question is if LAMMPS is limited to calculate high viscosities?
This is not a LAMMPS limitation, but rather a limitation of molecular dynamics in general. To measure very high viscosity using equilibrium measurement, long duration must be simulated, which is difficult to do with a small timestep.
Here, how did you choose the duration of your simulation?
This is obviously an issue. Higher viscosity means slower dynamics and longer characteristic times, so the best parameters for the correlation function cannot be the same as for a low viscosity system.
You should increase simulation time, but not necessarily a timestep. Higher timestep lowers the accuracy of numerical integration and therefore you will soon produce an unstable simulation which will crash.
In any case, a 100 fs simulation is nothing. The relevant simulations lengths in MD simulations range from several ps up to hundreds of ns.