How to heat nanowire and then cool it down

Hi All,

I am doing some silicon nanowire bending simulations to see how bending affects electronic properties. So far I have been doing just energy minimization (no MD was used). Now, for verification purpose, I would like to do some MD.

This is what I want to do:
Take a silicon nanowire (I read coordinates from a data file), heat it to about 1000K and cool down to lowest temperature possible. The purpose is to see if there is significant change in xyz positions of atoms compared to when just minimization is done. I have been trying simple nvt fix like,

velocity all create 001.0 12345 mom yes rot no

fix 1 all nvt temp 001.0 1000.0 100.0
run 10000
unfix 1

fix 1 all nvt temp 1000.0 001.0 100.0
run 10000

but it seems the temperature does not increase beyond a few tenths of a Kelvin (before the fix I zero out velocities of some ends atoms to keep the nanowire fixed in position) in the first run.

Can you please suggest some simple steps to achieving my goal here?


You didn't say what units you are using. Read
about the importance of the final 100.0 value in
your fix nvt command on the fix nvt doc page.
You may be doing next to nothing in 10,000 steps.


Thanks for your reply.

I am using metal units.

And yes, I found that 100.0 was too long a value. The temperature behavior seems fine with a Tdamp value of 0.1. But I see my crystalline starting structure becomes amorphous (except for the fixed end atom layers) as temperature is increased to 800K, and it does not recrystallize as temperature is cooled down. Am I doing it too fast?

What is the best way to optimize a crystalline (nanowire) structure with MD?


I don't know. I'd find a paper that does it and try to reproduce that
results first.


Maybe the timestep size was too large, and maybe the pair_style used
was not able to reasonable describe Si nanowires.


Thanks Ray.

The time step problem is now solved and the nvt works for bulk silicon. But I am still trying to get it to work for my nanowire.
My requirement is this:
Take a nanowire (straight or bent) with hexagonal cross section (I read in data file listing box size and atomic positions), fix a few layers of atoms at both ends and do MD to get optimized atomic positions. I want the free atoms to relax but the nanowire should remain more or less crystalline, meaning the bonds will not distort that much. So far I see that as I heat the system the nanowire portion that is free to relax just twists and jumbles and does not recrystallize as I cool down the system. I have done both heating and cooling for up to 200ps, and tried both periodic and fixed boundaries. I understand there will be larger change in positions of surface atoms as passivation is missing. Am I doing/asking for something unreasonable?

Thanks for your help.

Hi Golam,

Since bulk silicon and Si nanowire are too far disimilar from each
other, a potential (also true for time step size) that works for bulk
silicon does not necessarily work for Si nanowire. This so-called
transferability issue is the weakness of empirical potentials. Try
NVE first to verify that reasonable energy/temperature profiles and
stable structure are obtained. If NVE fails with a timestep as small
as 0.01 ps, then it most likely indicates the incapability of the
potential used to describe Si nanowire.

It seems you have a nanowire in a big box. Are you box dimensions
large enough so that the wire does not interact with its periodic
image? If you have fixed ends, due to thermal expansion the middle
free portion of your nanowire can only "twist and jumble" since it
couldn't expand. Also, if you have NVT applied to the middle portion,
then it really is not "free" (free would imply NVE). If the nanowire
does not recrystallize upon cooling, it may suggest that it has
evolved to another local minima on the potential energy surface upon
heating so it does not relax back to the original local minima you
started with.