Towards a statistically validated metrology procedure for assessing position accuracy of edge-emitting lasers to photonic integrated circuits

Abstract Hybrid integration of edge-emitting lasers (EELs) onto silicon nitride photonic integrated circuits (PICs) requires micrometre-scale placement with sub-micrometre verification and a reliable, non-destructive way to verify alignment after assembly. An Infrared (IR)-readable Nano-Vernier metrology is introduced that quantifies lateral offsets at multiple die corners and combines them with a rigid-body fit for statistical evaluation. Using GaAs surrogate dies bonded with Sn-Ag (tin-silver) solder onto oxide-capped DRIE (deep reactive ion etching) pillars, three assembly configurations were studied: two stops, one stop, and no stops. Across all assemblies, residual misalignments with two stops were tightly controlled (x: 1.57 µm ±1.26 µm; y: 0.39 µm ±1.04 µm), while removing the y-stop increased y variability (0.95 µm ±1.60 µm). Without stops, spreads grew substantially (x: 7.82 µm ±3.35 µm; y: 5.02 µm ±2.09 µm). Paired comparisons confirmed a consistent x > y bias (mean difference = + 1.12 µm; 95% CI [0.87, 1.37]; p = 1.40 × 10 ⁻¹⁵ by t -test; p = 1.52 × 10 ⁻¹⁹ by Wilcoxon). Each Vernier read carried a typical expanded uncertainty of U ≈ 0.15 µm to 0.22 µm ( k ≈ 2), far below process spreads. These results demonstrate sub-micrometre metrology with traceable uncertainties and tight, stop-controlled placement: in the two-stop case the y translation is sub-1 µm for the large majority of dies, whereas x exhibits a repeatable micrometre-scale bias that is therefore trimmable by stop/pad design; the penalties of reduced mechanical confinement are quantified by the one-stop and no-stop cohorts. The method is fast, non-destructive, and compatible with infrared microscopy, making it suitable for in-line statistical process control and design-of-experiments in hybrid photonic packaging.