Please copy to the mailing list. Comments and replies below.
Ray
Dr. Shan,
I have heard many times you and others say how each potential models
different parameters. I have noticed how each potential does indeed produce
slightly different values in lattice parameters, rdf's etc.
Each potential is fitted, or trained, to reproduce certain properties
of a material. Some potentials focus on some properties that others
put little emphasis on. It is very normal that each potential
generates different results and sometimes the extent of difference is
quite significant.
Can you give me some examples of when you might decide to use Tersoff over
Buckingham or vice versa. Also, might you use Morse? Could you tell me some
of the advantages of the Comb Potential that you have been working on? I
have been working with the Comb potential some and noticed that it does cost
a lot more as I re-calculate partial charges every 10 timesteps.
For deciding which potentials to use for your simulations, it is
always useful to consider the FATE rule, which was proposed by Prof.
Don Brenner and stands for flexibility, accuracy, transferability, and
efficiency. Is the potential flexible enough to model the properties
that you wish to model? Is the potential accurate in
reproducing/predicting these properties? Is the potential
transferrable enough to describe properties not considered in the
fitting? And consider the computational cost of the chosen potential.
These potentials you mentioned are very different from each other.
One is charge neutral, one has fixed charge, one is reactive, and one
is variable charge and reactive. To understand each of these
potentials, it is best to read their publications carefully.
When you say it is reactive, are you saying that it is useful for simulations that have chemical reactions? Like one element mixing with another? Or does it mean something else?
I understand that one is charge neutral, one has variable charge, and one has fixed charge. Amongst those three, could you give me examples where one potential might be used over another? I have read the papers but I still do not have a feeling of which potential might be used for different examples.
Can you think of an instance in which you say to yourself, “Tersoff would be better for this than buckingham,” or “Comb would model this property more effectively than a fixed charge like Buckingham would.”
Ben
When you say it is reactive, are you saying that it is useful for
simulations that have chemical reactions? Like one element mixing with
reactive means, it *can* model chemical reactions (change in bonding)
to some degree.
another? Or does it mean something else?
I understand that one is charge neutral, one has variable charge, and one
has fixed charge. Amongst those three, could you give me examples where one
potential might be used over another? I have read the papers but I still do
not have a feeling of which potential might be used for different examples.
Can you think of an instance in which you say to yourself, "Tersoff would be
better for this than buckingham," or "Comb would model this property more
effectively than a fixed charge like Buckingham would."
there is no simple rule for that. outside of reading what other people
have used if for and particularly paying attention to their
justifications as well as making experiments and tests yourself.
empirical potentials should *never* used without some validation
tests. those can be rather simple, but without validation any kind of
result would be tainted.
axel.