Structure-Based Machine Learning Enables Discovery of Mn4+-Activated Red-Light Fluorides for Ultrawide-Gamut Mini-Light-Emitting Diodes

 https://doi.org/10.1002/adfm.202313490

Mn⁴⁺-activated fluorides with a saturated red color and sharp line emission are ideal for applications in the light-emitting diodes (LEDs) backlight for displays. However, the emissions attributed to ²E→⁴A₂ parity and spin-forbidden transitions limit the design and adjustments of emission wavelength and chromaticity coordinates. Herein, machine learning algorithms are used to build a wavelength-prediction model for Mn⁴⁺-activated fluorides. The model precisely identifies the key structural features that affect wavelengths and discovers target materials. The predicted candidate Cs₂NaAlF₆:Mn⁴⁺ (CNAF) with a long-wavelength zero-phonon-line emission at 628 nm exhibits a redshift in comparison with other reported Mn⁴⁺-activated fluorides and commercial K₂SiF₆:Mn⁴⁺, but maintains narrow spectral emission with full-width half maximum (FWHM) of 11.2 nm. The redshift and narrow spectra result in a color purity of 99.7% and Commission Internationale de L'Eclairage (CIE) chromaticity coordinate of (0.7032,0.2967) that is close to the pure red-light point of Recommendation BT. 2020 (Rec. 2020). Moreover, CNAF is prepared as a transparent red-light film, and the device fabricated using the blue-light mini-LEDs, green quantum-dot film, and CNAF film exhibits a wide color-gamut of 121.5% National Television Standards Committee (NTSC) or 90.6% Rec. 2020, suggesting that CNAF has potential for wide-color-gamut displays.