ATKError: inverse(DZMatrix const &) : Could not LU factorize!

[Hasan Sahin]

During two-probe calculation ATK gives the error

ATKError: inverse(DZMatrix const &) : Could not LU factorize!


what does it mean?

[Nordland]

It is a bit hard to tell, but it means that it tried to invert a matrix, that could not be LU decompose (as the errors says).

1) What version of ATK are you using? This could perhaps give a clue on the what the error means ?
2) Do you have the script that produced the error ?
3) How does the geometry look like in VNL?

[Hasan Sahin]

thank you Nordland

1-) Latest version of ATK
2-) Inserting the geometry, script was generated by Nanoscripter (in former version of VNL)
3-) I controlled the geometry, it seems normal, there is no overlapping atoms.

Now I am trying to generate the scirpt with latest version of VNL.

[Nordland]

If you are using the new experimental version of ElectrodeConstraint (DensityMatrix),
then you will get an error (LU factorize) if the all charge leaves the system,
does your "q" goes to zero before the error?

Then I suggest tuning the parameters for the energy contour integral and the eigenstate occupation parameters.

[Hasan Sahin]

now it is ok. It is about electrode-device system geometry. The part of my electrode which does not match with the edge of device region causes this kind of problem.

[Nordland]

Great!