Is the `Finite Bias Anomaly' in planar GaAs-Superconductor junctons caused by point-contact like structures?

We correlate transmission electron microscope (TEM) pictures of superconducting In contacts to an AlGaAs/GaAs heterojunction with differential conductance spectroscopy performed on the same heterojunction. Metals deposited onto a (100) AlGaAs/GaAs heterostructure do not form planar contacts but, during thermal annealing, grow down into the heterostructure along crystallographic planes in pyramid-like `point contacts'. Random surface nucleation and growth gives rise to a different interface transmission for each superconducting point contact. Samples annealed for different times, and therefore having different contact geometry, show variations in $dI/dV$ characteristic of ballistic transport of Cooper pairs, wave interference between different point emitters, and different types of weak localization corrections to Giaever tunneling. We give a possible mechanism whereby the `finite bias anomaly' of Poirier et al. (Phys. Rev. Lett., {\bf 79}, 2105 (1997)), also observed in these samples, can arise by adding the conductance of independent superconducting point emitters in parallel.