FAQ

Failing tests

When starting to use icet it is a good idea to run all the tests to check that everything works as expected. A few common reasons why tests may fail include:

  • Out of sync test and source-code. The tests have to match the version of the package.

  • Old packages: If the numpy/sympy/spglib version that has been installed does not match the version requirements some tests can fail.

Does icet run on Linux/Mac/Windows?

icet is currently well tested only on recent OS versions in the Linux and Mac family, but icet should in principle work fine on Windows too. If you encounter any issues when running icet on Windows, please send an e-mail too icet@materialsmodeling.org and describe the problem.

icet crashes with the error Failed to find site by position (findLatticeSiteByPosition)

This error usually occurs when trying to calculate the cluster vector of a structure that is not a supercell of the primitive structure that the ClusterSpace is based on. Sometimes, the reason is too strict numerical tolerances, which is why the error message suggests changing the tolerances when creating the cluster space. If changing the tolerances does not help, we recommend the following:

  • Ensure that the input structure is sensible, i.e., that it matches the primitive structure of the cluster space.

  • Is the structure relaxed, such that the atoms are no longer at the ideal lattice sites? If yes, you need to map the structure back to the ideal cell. This can be accomplished with the function map_structure_to_reference.

  • Occasionally, wrapping your structure may do the trick.

Does icet support 2D systems?

Structures in icet should always have periodic boundary conditions in all three directions. What you can still do, however, is to add sufficient vacuum in one direction such that the cutoffs in your cluster space do not extend from one periodic image to the next in that direction. For more information, see here.

Does icet support arbitrary point group symmetries?

No, icet currently only supports crystallographic point group symmetries. The symmetries of, for example, an icosahedral nanoparticle will thus not be properly handled in icet.

The SQS generator gives me a weird supercell

The function generate_sqs searches for the optimal SQS (special quasirandom structure) not only by decorating a supercell with a fixed shape, but also by trying different supercells. Sometimes the optimal SQS has a cell shape that looks quite different from the input primitive cell, but it is always a supercell of the primitive structure, i.e., it has the same symmetry etc.

In some cases, it can be a good idea to generate SQS with a specific supercell. To this end, there is a separate function, generate_sqs_from_supercell, to which the user can specify among which supercell(s) the SQS should be sought.

The structure mapping function fails

The structure mapping function, map_structure_to_reference, sometimes fails if the structure to be mapped differs too much from the ideal cell. A particularly important aspect is that the structure to be mapped cannot be significantly rotated from the primitive cell. Any major rotation has to be reverted by the user prior to mapping.