TS: Stop if equilibrium integration contour does not include lowest eigenvalue.
We've been stuck for about 3 years trying to converge a system with Pt in TranSIESTA. We tried many things and it didn't converge, so we gave up.
It now seems like the simple reason it was not converging is because we were using a contour that only went to -40eV, while our hamiltonian had eigenvalues as low as -97eV. Changing the contour to reach -110eV has made the system converge flawlessly.
Since probably almost everyone is using the same template for TS inputs, I assume that there are or there will be in the future other people that face this problem. If we (including Pablo) didn't think of this, probably most users will not think of this.
To avoid incorrect calculations and headaches for users, I think TranSIESTA calculations should fail if the contour does not englobe all eigenvalues. Discussing this with @garalb, we think this could be checked on the 0V TranSIESTA run, when the initialization is done with SIESTA. You already have the eigenvalues from SIESTA, so you could check if the lowest one is included in the contours. If not, SIESTA should die with a meaningful error message.
If you agree on this approach or propose another approach I can create the MR.