I check the displacement of the particles at the first step. Also the force acting on the probe particle shows it is in glassy phase.
yes, I tried 4000 particles and even 10000 and 20000.
Yes, I understand that your temperature is reducing due to the conversion of Ep <=> Ek. How big is the decay? Could this be triggering the glass transition? Because note that the āpreparation stageā that was skipped probably creates an initial configuration that has the due Ep to run the dynamics in the nve ensemble and keep the value of the reduced temperature of 0.466. In your case, you dont have an initial microstate that is proper to do that (I doubt that the energy minimization is doing this for you), which is actually why your average temperature changes (decays, as you told us) during the nve dynamics. So it could be that the people that did this work originally did not have a temperature decay when running in the nve ensemble.
I would save a dump file during the NVE part and see if the glass transition is happening there.
It decays to lower temperatures just in a few time steps.
Yes, you are right. It goes to a temperature lower than the glass transition in the nve stage. However, during the nvt stage, it goes back to the higher temperature 0.466 which is supposed to be liquid and has no memory. Is not that right?
Hmmm not necessarily. Phase transitions in general require some āinitiationā event (related to an activation energy) that can take a while to happen during the dynamics. For example, you can have a liquid below the melting temperature and a solid above the melting temperature. Ofc there is a limitation for the degree of undercooling and overheating that you can achieve for each system as the activation energy for the energy barrier for the āinitiation eventā becomes smaller and smaller and thus the probability for the phase transition event becomes larger and larger. In these overheated and undercooled scrnarios, the initial phase is no longer the thermodynamic stable one but it takes some time for the dynamics to lead to the āinitiation eventā. This is specially critical in simulations, where we are confined to scales different than the real life one. At least thats how it goes for usual solid (xtalline)/liquid. I suppose for glass transition is the same story (maybe it is a good idea to check it in some book tho), if my thermodynamics of phase transition is not too rusty.
So maybe it is solidifying during the nve due to the drop of temperature and remaining solid even at the high temperature because of lack of time. Try saving a dump file and assessing this further to see if itās the case.
If it is the case, you can think of dropping the nve unless you start from an initial configuration that has a suitable initial potential energy to keep more or less the given value of temperature (0.466) throughout the nve dynamics.
It seems It is not the case!
I removed the nve stages and just kept nvt integration. The minimization can not be completely removed because the interaction of the particles leads to an explosion. However, I reduced the number of minimization steps to a few ones to make sure It does not explode but remains in rather high energy providing T>0.466. Then using a nvt the current temperature decreases to T=0.466, and by another nvt the system evaluates at T=0.466. However, the issue is still there!
Hmmā¦ :///
Then I fear I am out of ideas of what could be the issue.
Unless the initial configuration is somehow propense to lead to the amorphization (so the preparation would be useful) or if you are extremely unlucky and getting it to happen in the very first steps of the nvt where the temperature fluctuates a lot (given that it is not really an initial microstate corresponding to the thermodynamic state you are studying Iām the nvt). But I donāt even know even know if these things are actually possibleā¦ especially the last oneā¦
I suppose that to give more ideas of what could possibly be the issue I would need to know better your topic and/or have actual time to read somewhat carefully the paper you linked, but the truth is that I donāt ://
In any case, good luck solving the issue!! ^^
1 Like
What if you started by initially placing the particles on a lattice instead of randomly? The crystal->liquid transition far enough above the melting point should not go through a glassy phase (I am very bad with this stuff so I am simply expressing a hopeful thought).
2 Likes