Author Topic: Problem with identifying V. Whole surface interpreted as V by WS modifier  (Read 1020 times)


  • Jr. Member
  • **
  • Posts: 68
Hi all,

I am trying to identify interstitials and vacancies after a collision cascade in Fe. This is done with LAMMPS code. A projectile hits an open surface with a kinetic energy of 10 keV and the evolution of the system is followed during approx 30 ps.
I have implemented a script with the Wigner-Seitz modifier to identify SIA and V, it works well.
However, I observed that in some cases, the number of vacancies or interstitials is extremely high, very far from the expected mean. For instance, if I expect in average 100 SIA to form, the script gives me 3000. After digging a bit, I found that this happens because in some cases, the WS modifier interprets one of the surfaces as made of V and the one in front as SIA (see picture attached).
I thus have to remove these surface atoms from the WS analysis.
After discussing it with Alex in a previous post, he recommended identifying surface atoms with the CoordinationNumber modifier. In BCC crystal like Fe, the coordination number of atoms in bulk is 8. Thus, surface atoms should have a lower coordination number.
Using the CoordinationNumber modifier I can indeed identify surface atoms. With 'Coordination == 4' in the SelectExpression modifier I can isolate atoms from both surfaces that induce wrong results. However, I found that some of the defects also have a CN=4. Thus, if I remove atoms from the surface by selecting them with 'Coordination == 4' I also remove some of the "real" defects.

Why WS interprets surface atoms as defects? To answer this question, the only idea I can think of is that during the evolution of the system, the system expands a little bit since the surface is open and the temperature increases. Thus, when the WS modifier tries to identify SIA and V by comparing the reference crystal and the crystal with defects, it sees that the surface atoms have moved and interprets they are defects. Does it make sense?

How can I separate real defects from surface atoms without removing "real" defects with the CN? Any idea?

Many thanks in advance for any help.
Best regards,

Alexander Stukowski

  • Administrator
  • Hero Member
  • *****
  • Posts: 638
Hi Christophe,

I am wondering what you use as reference configuration for the WS analysis? Do you use the ideal lattice before relaxation or after?
The surface relaxation likely leads to a shift of atomic planes. Connected to this is the question whether the artefacts on the surface occur already before radiation or only after?

As you have already noted, the defects produced during the cascade can lead to an expansion of the crystal (elastic, thermal or both), which shifts all atomic positions. This lattice strain can lead to artefacts in the output of the WS analysis. The problem -and a possible solution- are discussed on page 2 of the following article:

However, this is not something that can be readily done with Ovito. For the above publication we developed a special post-processing tool which removed the elastic displacements from the atomic positions. After this 'cleaning' step, the configuration could be processed with the WS analysis to find point defects.


  • Jr. Member
  • **
  • Posts: 68
The reference configuration that I use is the crystal after equilibration, before irradiation.

But, how to know if equilibration is correctly done? Equilibration is a tricky point. There many posts on that in LAMMPS forum. The way I do it was suggested by Axel Kohlmeyer:

- Thermostat of 2 layers at 1K (or 300K) around the crystal except for the open surface
- Initialization of energies of atoms at T
- Minimization of potential energy
- Equilibration during 2.5 ps with Langevin with short time constant
- Equilibrationduring 2.5 ps with NVT with longer time constant

Either equilibration is not properly done, or there is an additional expansion during cascade or both.

Regarding equilibration, I guess I could leave more time to ensure that all the expansion that should occur takes place before irradiation.
If there is an expansion during cascade, then I guess I'll have to use the method you propose.

Thanks again and best regards,