How to determine the onset of melting using the RDF

I have simulated a gold super cube of 4000 atoms, from a temperature of 1 to 3000K, using an NVE ensemble and Berendsen Thermostat. The temperature linearly increases over the 100 ps simulation and the gold cube begins to melt.
My question is, how can the onset of melting be determined using the RDF function?
It is evident from the visualization and from the RDF itself, that long-range order is lost, however, it is difficult to define an exact temperature where melting could be said to occur.

Are there any rules or definitions related to the onset of melting or the melting point using the RDF?

If I understand your description correctly, you heated your sample from 1 to 3000 K in 100 ps. That’s way too quick to get any meaningful result. There is zero statistical sampling in the RDF you can compute at any intermediate temperature. Also, I expect large hysteresis effects.

I agree with @hothello, 100 ps is way too short. Also, I remember that more advanced methods are usually used to determine the melting temperature of metals, such as probing the velocity of the liquid/solid interface in a partially melted system. The problem with the method you are using is that there is no guaranty that the melting will occur at the melting temperature.

This is a very bad choice of thermostat for this purpose.

The RDF is a bad indicator for detecting melting. It is rather insensitive.

But more importantly, melting is an activated process, so you’ll have to deal with melting not happening instantaneously, but rather depending on a sufficiently large fluctuation happening that crosses the activation barrier. This makes any study using a temperature ramp on a single phase system subject to hysteresis effects (ever heard of supercooled liquids? this is the opposite).

The most commonly used procedure to determine a melting point from simulations is the coexistence method where you monitor the phase growth or lack thereof of half liquid half solid system. This has been discussed repeatedly in this forum, and there should be plenty information about it in the published literature.

1 Like