Is there any particular reason why you run this a spin-polarized? I don't see any reason to expect any spin-dependence in this system, none of the elements involved should exhibit any magnetic behavior. Although I may be wrong of course... But perhaps it's at least faster to try different convergence tests without spin, as it reduces time-consumption.
The structure is reasonable (heavy, of course), but it's worth noticing that the Au inside the central region is a bit thin. If you have 3 layers in the electrode, I always make sure to have at least one more in the central region, so 4 in your case.
Finally, the SingleZeta basis sets are often really not useful, and having it for Gold here is probably the primary reason for the bad convergence. I understand SingleZetaPolarized will increase the computation time quite a bit, but it is most likely necessary.
