Machine-learning identified nitride quantum wells for enhanced red emission

III-nitride heterostructures are critical for achieving polychromatic micron-scale pixels in extended-reality and biomedical applications, however red emission remains challenging. Here, we address this challenge by systematically exploring their design space with machine learning. We surprisingly uncover that larger polarization fields globally correlate with higher wave-function overlaps for red emission. The key is the quantum-confined Stark effect, which enables a reduction in well width while avoiding the associated increase in In composition. We outline strategies for engineering polarization fields through the tuning of composition, strain, and interlayer and quantum-barrier thicknesses. Without polarization fields, efficient red emission would not be possible in nitride LEDs given the current challenges in nitride epitaxy. Our results show how polarization can improve light emission from polar semiconductor heterostructures.