Dear Sergey,

Indeed, the automatic routine that does the attachment can only work on one unique lattice.
There are many solutions to your problem. 

- You can work on 3D cubic lattice. (that's probably the easiest). Note that Kwant works in arbitrary dimension so you can
also embed a 3D problem in a 4D one.

- You can attatch your 1D lead to a single site 1D system and then add a hopping between this site and your 2D system.

Best regards,
Xavier

Le 14 avr. 2015 à 02:00, Sergey <sereza@gmail.com> a écrit :

Dear Colleagues,
  I want to measure the decay of wave-function in magnetic field.  For that I create a 2D shape
and want to attach a 1D lead to only one internal lattice site of a 2D shape. (imagine, that this 1D chain
is pointing in the third direction, although this embedding does not matter for physics)  
Does this make sense?
  There is a technical problem that 1D families differ from 2D ones and I get a message:
"ValueError: Sites with site families (kwant.lattice.Monatomic([[1.0]], [0.0], ''),) do not appear in the system, hence the system does not interrupt the lead."
Any hints?

Thank you,
Sergey