Bulk energy has thermal and system size fluctuations?

Hi,
      I am running MEAM potential for pure bcc-U at 400K and observing the following:

1. For same starting structure and similar simulation conditions, bulk energy shows variation between different runs (-10395.3 eV, -10396.2 eV, -10395.4 eV for 3456 atoms). I guess this is because of thermal fluctuations as similar runs produced exactly same energy at 1K (-10793.8 eV for all 3 runs).

2. As I change the system size (2000 atoms, 3456 atoms, 6750 atoms), energy/atom is increasing. At 1K (minimal or no thermal fluctuations: 2000 atoms gave -3.09414 eV/atom and 3456 atoms gave -3.09717 eV/atom). Is it some kind of rounding off error or I am doing something wrong in my simulations?

I am enclosing my compilation and input script with this. Please let me know whether I am missing something or observing a normal simulation behavior.

Hi,
I am running MEAM potential for pure bcc-U at 400K and observing
the following:

1. For same starting structure and similar simulation conditions, bulk
energy shows variation between different runs (-10395.3 eV, -10396.2 eV,
-10395.4 eV for 3456 atoms). I guess this is because of thermal
fluctuations as similar runs produced exactly same energy at 1K
(-10793.8 eV for all 3 runs).

it may also depend on the nature of the potential and whether its
energy remains continuous when atoms enter or leave the cutoff,
and of course the time it needs to equilibrate. 50,000 steps is not a lot.

2. As I change the system size (2000 atoms, 3456 atoms, 6750 atoms),
energy/atom is increasing. At 1K (minimal or no thermal fluctuations:
2000 atoms gave -3.09414 eV/atom and 3456 atoms gave -3.09717 eV/atom).
Is it some kind of rounding off error or I am doing something wrong in
my simulations?

it may also depend on your initial structure and how well it
matches the box size. hard to tell without taking a closer look.
this is something that you have confirm to yourself.

axel.