Deep Learning for Chromatic Optimization in Dual-Color Mini-LEDs for Aircraft Displays

GaN-based miniaturized light-emitting diodes (mini-LEDs) have emerged as highly promising light sources for high-performance aircraft cockpit displays. Among these diodes, vertically stacked blue‒green dual-color mini-LEDs are fabricated and show excellent color-tunable properties. When packaged with K₂SiF₆:Mn⁴⁺, an impressive color gamut area ratio of 113.63% NTSC is demonstrated. However, vertical optical crosstalk originates from the photoluminescence (PL) effect of the green epitaxial layer when the blue mini-LED is powered on, making precise chromatic characteristics difficult to obtain. To address this problem, a deep neural network (DNN) is proposed, which combines forward modeling and inverse design in a tandem architecture. This DNN supports a current modulation scheme that enables precise control of chromaticity coordinates, achieving a Δu′v′ of 0.003. These advancements in materials and device strategies pave the way for developing low color difference, high-resolution display systems for aircraft cockpits.