ISF of Si

LAMMPS LAMMPS Mailing List Mirror
1

Dear All,

There was a problem when I calculated the intrinsic stacking fault (ISF) of Si. Two different potentials, namely, Tersoff (J. Tersoff, PRB, 37, 6991 (1988)) and SW (Stillinger and T. Weber, Phys. Rev. B, 31, 5262, (1985)), were used. Different simulation boxes were tested. The ISF was finally obtained as 0.112 eV/A2 for Tersoff and 0.096 eV/A2 for SW, both are much greater than that of DFT method with a result of 0.006 eV/A2 (Phys. Rev. B, 50, 5890, 1994 ). I tried to solve this problem, but I failed at last.

I set the orients x [112], y [-110], z [-1-11], the displacement of a half relative to another bp= a/6[112]= 5.4308 A. My code is as follows,

------------------------ INITIALIZATION ----------------------

units metal
dimension 3
boundary p p s
atom_style atomic
variable latparam1 equal 5.4308

variable xdim equal {latparam1}*sqrt(6)/2*30 variable ydim equal {latparam1}sqrt(2)/230

----------------------- ATOM DEFINITION ----------------------

lattice diamond ${latparam1}

region 1 block 0 {xdim} 0 {ydim} 0 100
region 2 block 0 {xdim} 0 {ydim} 100 200
region whole block 0 {xdim} 0 {ydim} 0 100 units box

create_box 1 whole

lattice diamond ${latparam1} orient x 1 1 2 orient y -1 1 0 orient z -1 -1 1
create_atoms 1 region 1

lattice diamond ${latparam1} orient x 1 1 2 orient y -1 1 0 orient z -1 -1 1
create_atoms 1 region 2

----------------------- FORCE FIELDS -----------------------

pair_style tersoff
pair_coeff * * Si.tersoff Si

------------------------- SETTINGS --------------------------

group top region 1
group bot region 2

------------------------- Displacement -----------------------

displace_atoms bot move -2.2171 0.0 0.0 units box
compute peratom all pe/atom
compute eatoms all reduce sum c_peratom

Dump to comp for Ovito post processing

dump 1 all custom 1 dump.comp.* id type xs ys zs c_peratom fx fy fz

thermo 1
thermo_style custom step pe c_eatoms

mass * 1.0

fix 1 all setforce 0 0 NULL

min_style cg
minimize 1e-10 1e-10 1 1

SIMULATION DONE

print ${x}

print “All done”

----------------------- END --------------------

Are there any mistakes in my code for the large discrepancy?

Thank
in advance.

Sincerely,

-Wilhelm

2

Dear All,

There was a problem when I calculated the intrinsic stacking fault (ISF) of
Si. Two different potentials, namely, Tersoff (J. Tersoff, PRB, 37, 6991
(1988)) and SW (Stillinger and T. Weber, Phys. Rev. B, 31, 5262, (1985)),
were used. Different simulation boxes were tested. The ISF was finally
obtained as 0.112 eV/A2 for Tersoff and 0.096 eV/A2 for SW, both are much
greater than that of DFT method with a result of 0.006 eV/A2 (Phys. Rev. B,
50, 5890, 1994 ). I tried to solve this problem, but I failed at last.

are you insinuating that it is a problem, that empirical potentials
give a different result than a DFT calculation (and a pretty old DFT
calculation at that which you compare to potentials that are even
older)? do you have any evidence they *should* be closer? have you
validated the DFT calculation?

axel.

3

Dear Axel,

Thanks for your reply.

The experimental ISF of Si is 50~70 mJ/m2, and calculated ISF is 33~190 mJ/m2 (PRB, 32, 7979, 1985). However, my calculataion is 1536~1792 mJ/m2. The empirical potentials should not give out a result much larger than “old” results. I try to figure out the reason for the discrepancy. As a new lammps user, I am not very sure about my code so I guess maybe there some “mistakes” in my code?

Sincerely,

-Wilhelm

4

Dear Axel,

Thanks for your reply.

The experimental ISF of Si is 50~70 mJ/m2, and calculated ISF is 33~190
mJ/m2 (PRB, 32, 7979, 1985). However, my calculataion is 1536~1792 mJ/m2.
The empirical potentials should not give out a result much larger than "old"
results. I try to figure out the reason for the discrepancy. As a new lammps

so why didn't you provide *this* explanation in the first place? the
better and the more to the point your explanation of your problem, the
larger the chances you get a useful answer. if you lead people to the
wrong conclusion, it is your own fault.

user, I am not very sure about my code so I guess maybe there some
"mistakes" in my code?

well, unless you get very lucky and somebody has done the same kind of
calculation before and immediately spots some obvious error, there are
very little chances that you will get somebody to help you.

since this is a rather old paper, there should be fairly detailed
explanations of how the simulation was set up (nowadays a lot of the
detail is skipped over, sometimes too much), so you can follow those
directions. there may be other papers comparing to it or referring to
it with more detail. so there is an opportunity to find more details.

axel.

5

Could the obvious error be that the max number of iterations and force
evaluations for minimization are one?

6

Dear Trautt,

Thanks for your reply. I adjusted the max number of iterations and got no much change of calculated ISF. But your suggestion of increasing the iteration times is valuableto me. Return to my problem, after inspection of the disturbed lattice with a relative displacement of a[112]/6, I found the reason for the large ISF is that the relative rigid shift between the two halves is in the shuffle plane. Then I modified the codes to make the relative shift in the glide plane, which can lead to a reasonable result.

Best regards,

-Wilhelm