Dear All,
I tried to calculate the melting point for the Silicon bulk with the parameters bellow but it gave me a wrong value for the melting point(about 2000) , but the right value (literature) is about 1600.
please,whre is the problem, in the input file ?
Dear All,
I tried to calculate the melting point for the Silicon bulk with the
parameters bellow but it gave me a wrong value for the melting point(about
2000) , but the right value (literature) is about 1600.
does "literature" here mean the experimental literature, or the paper
describing the potential you are using?
please,whre is the problem, in the input file ?
the problem is in your thinking.
have you tried the same simulation with a different number of
timesteps, i.e. a different rate of heating?
have you tried the same simulation starting from a high temperature
and recording the point where the system solidifies?
have you tried the same two simulations with different system sizes?
if you do this, you should learn a few things, that are well
understood and documented in the literature. it boils down to the
fact, that activated processes on the atomic scale are different than
on the macroscopic scale, so you cannot apply the same methodology. it
has been discussed on this very mailing list, that the most reliable
way to determine the melting point at the atomic scale is to do a
coexistence simulation where you monitor the size of the liquid and
the solid phase. if they remain the same, and not one is growing at
the expense of the other, you have the melting point.
axel.
Maybe you can try different potentials. From my knowledge, maybe SW potential would be better for melting for silicon. Hope this can be useful to you. Or you can try reaxff potential.
Maybe you can try different potentials. From my knowledge, maybe SW
potential would be better for melting for silicon. Hope this can be useful
to you. Or you can try reaxff potential.
this won't really help much, because the method of determining the
melting point from raising the temperature and observing energies is
flawed. it is dependent on system size, how quickly you raise the
temperature, the type of thermostat, and whether you raise or lower
the temperature.
if you shop around until you find a potential, that gives the desired
result for the given conditions, you just are getting the desired
answer for the wrong reason.
axel.
Thanks, Axel. Yes, you are totally correct. The melting can be dependent on the size and setup of the simulations. So either you can try to change the size or setup. Or you can find some paper which has those data for different potential. Thanks.
Thanks, Axel. Yes, you are totally correct. The melting can be dependent on
the size and setup of the simulations. So either you can try to change the
size or setup. Or you can find some paper which has those data for different
potential. Thanks.
or you determine it "the right way (tm)", e.g. from a set of
coexistence simulation at various temperature points around the
expected melting point.
like i was mentioning before, we have this issue being discussed on
this mailing list regularly, since there are always people that need
to be reminded of the implications of the difference in time and
length scales between computer simulations and experiments.
axel.
Yes, the “only” proper way of estimating the melting point is by constructing a two-layered system with one layer being amorphous (or liquid) and the other being crystal. Then run this coexistent structure at different temperatures. If you see for a certain temperature the solid-liquid line extends into the liquid region then this temperature is below the melting point. On the contrary if the solid-liquid line extends into the solid region then the temperature is above the m.p. A series of temperatures would allow you to bracket the m.p.
Estimating m.p. using energy vs. temperature or volume vs. temperature curves are extremely inaccurate - regardless of the potentials used.
Ray