There are several things that I would suggest to modify in your script.
- The extensions of your device structure appear to be small and need to be enlarged, especially for the left electrode. I enclose an image showing what the structure should look like in a decent setup. In general, there must be a sufficient amount of buffer layers between the actual semi-infinite left (right) electrode and the central region, see
http://docs.quantumwise.com/tutorials/atk_transport_calculations/atk_transport_calculations.html?highlight=atk%20transport.
- In the present script, you have the k-point mesh (6,6,10) and (6,6,1) for the left and right electrode, respectively. The k-point mesh should be consistent for the left and right Pd electrodes, e.g., set it to k_point_sampling=(6, 6, 101). You need a lot of k-points in the transport direction to properly match the Fermi level in the electrodes and central region, see
http://docs.quantumwise.com/tutorials/transport_kpoints/transport_kpoints.html?highlight=why%20point .
- The electron temperature might need to be increased to something like 1200 K (which corresponds to broadening of ~0.1 eV) since the system is metallic, see
http://docs.quantumwise.com/tutorials/fe_mgo_fe/fe_mgo_fe.html?highlight=initial%20stateIn addition, I would suggest first doing geometry optimization for the corresponding slab structure, allowing Co and some neighboring Pd atoms/layers to relax and setting constraints for all the other atoms, e.g., see the tutotial
http://docs.quantumwise.com/tutorials/geometry_optimization/geometry_optimization.html?highlight=geometry%20optimization on rigid and fixed constraint options. This will also allow you optimizing the separation distance between the electrodes.
You will then have a good initial structure for device calculation, and perhaps not even need to do geometry optimization for the device configuration afterwards. But you may do it indeed, if needed. To optimize the structure, you may just do it with LDA, this will save you the computational time.
