Toward Scalable 3D IRFPA Design: Optimized Meshing and Structural Stress Control in MoDiBI-Integrated MCT-Based IRFPAs

The drive for tactical and strategic overmatches requires advanced infrared focal plane arrays (IRFPAs) featuring megapixel formats and smaller pixel pitches. These high-performance arrays typically employ mercury cadmium telluride (MCT) detectors for their high quantum efficiency and tunable bandgap across the IR spectrum. However, the current heterointegration method, which uses indium bumps for electrical, thermal, and mechanical connection between the MCT detector and the Si readout integrated circuit (ROIC), faces a critical challenge. Thermal cycling induces significant localized stress around these bumps, causing interfacial delamination or localized fatigue failure. Overcoming this stress-induced vulnerability is paramount for evolving IRFPA reliability. Ensuring device reliability necessitates a robust thermomechanical optimization workflow, balancing model complexity and execution time. To address this, prior work established a framework for 2D and 3D finite-element (FE) simulations of packaged IRFPAs. Building on this, this submission utilizes multiparameter Monte Carlo tree search (MCTS) optimization on a 3D FE model, containing the maximum tractable number of pixels, to minimize the von Mises stress. Furthermore, the optimized geometries of the modified direct bond interconnect (MoDiBI) balanced composite structure (BCS) obtained from the 3D FE MCTS workflow are compared against 2D FE simulation results. This comparison determines the relative ranking errors and explores the predictive accuracy trade-offs associated with using reduced-dimensional or partially modeled 3D simulations.