The Selena project searches for neutrinoless double beta decay (0νββ) in ⁸²Se using a hybrid amorphous selenium (aSe) / CMOS pixel detector, combining a high Q-value (3 MeV) that sits above most natural backgrounds with the fine spatial and temporal resolution of CMOS readout for event classification and decay-chain tagging. This talk covers two threads of recent work. First, Geant4 simulations of beta-induced charge tracks in aSe were used to predict detector energy resolution by modeling ionization, drift, and recombination. An early anomaly in the simulated charge spectrum was traced to how the Penelope physics model treats where switching to the Livermore physics model resolved the artifact and improved projected resolution from 1.8% to 0.5%. Then an implementation of graph neural network exploiting spatial correlations between charge voxels gave a further modest gain (0.4%) but introduced a small reconstruction bias still to be resolved.

Second, I present progress on chip packaging, where conventional wire bonding and flip-chip methods both distort the electric field near the sensor edges or give unreliable high-voltage contact. As an alternative, we tested micro-extrusion printing (XTPL) of conductive silver traces across the resin gap between chip and PCB. Early attempts with a narrow 20 µm trace gave poor, high-resistance contact, but widening the trace to ~30 µm produced a much more reliable connection with resistance dropping to tens of ohms.