Did the scale factor have such huge impact on the elastic constant calculation?

Hi all,
I have changed the script to calculated the elastic constant of FCC silver crystal.
The “potential.mod" is as the following:

pair_style eam/fs
pair_coeff * * Ag.set Ag
neighbor 0.3 bin
min_style cg
min_modify dmax ${dmax} line quadratic
thermo 1
thermo_style custom step temp pe press pxx pyy pzz pxy pxz pyz lx ly lz vol
thermo_modify norm no
The “init.mod" is as the following:

variable a equal 4.07
lattice fcc $a
region box block 0 4.0 0 6.0 0 8.0
create_box 1 box
create_atoms 1 box
change_box all triclinic

Need to set mass to something, just to satisfy LAMMPS

mass 1 1.0e-20

I got the C11=25.06Gpa, C12=77.74Gpa and C44=34.28Gpa. However, if I change the box as: region box block 0 2.0 0 3.0 0 4.0 I got a totally different result. C11=-72.76Gpa, C12=62.55Gpa and C44=23.42Gpa.
How can I fix this problem? Thanks for your help.

P.S. I have checked my outfile, the starting energies of both cells are the same. The outfile are in the attachment.

in.elastic-2-3-4.log (37.3 KB)

in.elastic-4-6-8.log (44.7 KB)

Elastic constants are heavily affected by the finite size effect. You can not have structures that are too small. You should also verify the size effect by gradually increasing the system size until the elastic constants converge.