List of bugs
As any non-trivial software, FLEUR contains bugs. You can check the git history as well as the issues (open and closed) to see the bugs we are aware of.
While most bugs are non-critical and affect only special features or specific machines, some prevent you from doing your work and others might lead to wrong results. Most critical are of course bugs that lead to results that are incorrect but not so far off that the presence of the bug is obvious. Whenever we encounter such a bug, we will try to descibe it here to enable you to check your results.
If you encounter a bug, we strongly urge you not to simply discard the code but to contact us to help with improving FLEUR.
Known serious bugs
| Core-tail treatment wrong in non-collinear setups. | |
|---|---|
| Description | While the full core-tail expansion is incompatible with noco, even the simple treatment of the core tail charges done instead in the ctail="F" case is broken in the noco calculations. |
| Properties affected | As the charge is not correct, all quantities in a self-consistency are modified. |
| Significance | Low, only the magnetisation of the core-tail is not correctly transferred to the interstitial magnetisation. This effect should only be seen when investigating very small energy differences. It scales with the charge lost from the core as reported in the out file and hence is smaller if semicores are treated by LOs. |
| First affected version | very old bug, present in v26, probably dates back to the first noco implementations. |
| Status | Not yet fixed, use LOs to circumvent. |
| SOC and non-collinear magnetism cannot be used together. | |
|---|---|
| Decription | The combination of SOC and noco leads to wrong results. This, so called 'SOC in first variation' did not take the rotation of the local spin-frame into account correctly. |
| Properties affected | As the Hamiltonian is incorrect, energies as well as all other output quantities are wrong |
| Significance | High, the effect is as large as the SOC strength |
| First affected version | probably all MaX versions, at least since MaXR3 |
| Status | FIXED, bug should be fixed in MaXR5.1 |