How to obtain the internal stress of initial geometry

GULP
1

Dear Julian and all users,

I am thinking about how to obtain the static stress/pressure of my system with a force field, and comparing it with the results from lammps.

I saw keywords like ‘strain’ + ‘stress_out’ or assigning pressure to the system and obtaining the enthalpy instead of energy, but the former I think is for the MD in GULP, and the latter is to assign specific stress to the system.

I want to obtain the internal stress of the initial geometry with a specified force field, like the ‘thermo_style press’ used in lammps to print the initial pressure/stress.

what keywords in gulp I can use to calculate it.

thank you in advance.

wang

2

“stress_out” is the keyword to print out the effective pressure for a configuration. “atomic_stress” gives you the atom by atom breakdown. Neither of these are specific to MD & so can be used with “grad conp” (i.e. a single point gradient calculation).

3

Dear Julian,

Thanks for the help. Now I can successfully output the stress.

One thing that confused me was how Gulp gets the sum stress by summating each atomic stress. I guess it is obtained by multiplying coefficients, but not clear what it is.

Atomic stresses (GPa):

No. Atomic xx yy zz
Label /yz /xz /xy

1 O c 0.002607 -0.001930 -0.006458
0.008833 0.002583 -0.001451
2 Cu c 0.013078 0.007052 0.007052
-0.006597 0.000000 -0.000000
3 C1 c -0.009945 -0.009945 -0.061013
0.030007 0.030007 -0.020833
4 H c -0.013836 -0.013836 -0.051141
0.025335 0.025335 -0.013034
5 C2 c 0.018047 0.018047 0.014942
-0.009595 -0.009595 -0.007290
6 C3 c -0.008582 -0.008582 0.015828
-0.010061 -0.010061 0.018141

Sum -0.863881 -0.863881 -0.863881
0.000000 0.000000 0.000000

All the best,
wang

4

Interactions are just divided equally over all atoms involved in the interaction. This is the same for stress and site energies.

5

Dear Julian,

Thanks for your reply.

so I am thinking about the contribution to the sum stress from each type of atom also depends on the number of each atom type. For instance, the formula is O48Cu12C72H24, so the stress contribution from O is proportional to 0.002607*48 in stress_xx. is it correct?

wang

6

The atomic stress is an arbitrary partition of the total stress, but obviously it has to add up to the total & so must be proportional to the number of atoms.